scene_id string | label string | manifest string | manifest_text string | inquisitor_response string | cot_text string | debate_rounds int64 | final_verdict string | max_confidence float64 | region list | category_hit string | pattern_id int64 | system_prompt string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
fake_01792 | deepfake | {"objects": [{"name": "face", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_h... | Semantic Manifest:
Objects: face
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: t, e, m, p, o, r, a, l, _, s, h, i, f, t | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows rPPG_anomaly — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [59, 167, 499, 4... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows rPPG_anomaly — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [59, 167, 499, 4... | 1 | CONTRADICTION | 0.86 | [
59,
167,
499,
482
] | temporal_video | 514 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01476 | deepfake | {"objects": [{"name": "edge", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspective"... | Semantic Manifest:
Objects: edge, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 44 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.80
Regi... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 44 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.80
Regi... | 1 | CONTRADICTION | 0.8 | [
161,
68,
591,
475
] | frequency | 444 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01156 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}... | Semantic Manifest:
Objects: face
Lighting: overhead, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing face for specular reflection inconsistencies.
</observe>
<hypothesize>
The face shows abnormal specular-diffuse ratio. face shows specular highlight inconsistency with soft overhead lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern d... | <observe>
Analyzing face for specular reflection inconsistencies.
</observe>
<hypothesize>
The face shows abnormal specular-diffuse ratio. face shows specular highlight inconsistency with soft overhead lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern d... | 1 | CONTRADICTION | 0.83 | [
116,
106,
530,
491
] | specular_reflection | 100 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01543 | deepfake | {"objects": [{"name": "skin", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "persp... | Semantic Manifest:
Objects: skin, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 26 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.82
Regi... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 26 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.82
Regi... | 1 | CONTRADICTION | 0.82 | [
98,
51,
581,
490
] | frequency | 426 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01063 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["opposite_all"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "pers... | Semantic Manifest:
Objects: person, object
Lighting: top-left, shadows: ['opposite_all'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing lighting: top-left source with shadows at ['opposite_all'].
</observe>
<hypothesize>
This lighting_shadow pattern violates physical lighting laws. Single top-left source produces impossible shadow configuration: ['opposite_all']
</hypothesize>
<test>
Light at top-left cannot produce shadows in state... | <observe>
Analyzing lighting: top-left source with shadows at ['opposite_all'].
</observe>
<hypothesize>
This lighting_shadow pattern violates physical lighting laws. Single top-left source produces impossible shadow configuration: ['opposite_all']
</hypothesize>
<test>
Light at top-left cannot produce shadows in state... | 1 | CONTRADICTION | 0.95 | [
66,
188,
415,
317
] | lighting_shadow | 55 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00207 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.89 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00604 | clean | {"objects": [{"name": "person", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera... | Semantic Manifest:
Objects: person
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | 1 | CONSISTENT | 0.93 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01853 | deepfake | {"objects": [{"name": "hands", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_... | Semantic Manifest:
Objects: hands
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: f, r, a, m, e, _, r, a, t, e | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows shadow_temporal_shift — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [116, 1... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows shadow_temporal_shift — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [116, 1... | 1 | CONTRADICTION | 0.87 | [
116,
110,
524,
346
] | temporal_video | 521 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01116 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["sharp"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "pe... | Semantic Manifest:
Objects: face, background
Lighting: right, shadows: ['sharp'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Analyzing face for specular reflection inconsistencies.
</observe>
<hypothesize>
The face shows abnormal specular-diffuse ratio. Face shows abnormal specular-diffuse ratio for diffuse lighting with ['sharp']
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pa... | <observe>
Analyzing face for specular reflection inconsistencies.
</observe>
<hypothesize>
The face shows abnormal specular-diffuse ratio. Face shows abnormal specular-diffuse ratio for diffuse lighting with ['sharp']
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pa... | 1 | CONTRADICTION | 0.85 | [
194,
74,
535,
310
] | specular_reflection | 161 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00578 | clean | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.95 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01503 | deepfake | {"objects": [{"name": "skin", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: skin
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 41 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.81
Regi... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 41 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.81
Regi... | 1 | CONTRADICTION | 0.81 | [
70,
121,
435,
437
] | frequency | 441 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00585 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00789 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "per... | Semantic Manifest:
Objects: person, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | 1 | CONSISTENT | 0.95 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00987 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00017 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.89 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01850 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: t, e, m, p, o, r, a, l, _, s, h, i, f, t | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows rPPG_anomaly — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [69, 121, 464, 3... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows rPPG_anomaly — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [69, 121, 464, 3... | 1 | CONTRADICTION | 0.86 | [
69,
121,
464,
333
] | temporal_video | 514 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01221 | deepfake | {"objects": [{"name": "cheeks", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_h... | Semantic Manifest:
Objects: cheeks
Lighting: top-right, shadows: ['bottom'], intensity: soft
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Analyzing cheeks for specular reflection inconsistencies.
</observe>
<hypothesize>
The cheeks shows abnormal specular-diffuse ratio. cheeks shows specular highlight inconsistency with soft top-right lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pa... | <observe>
Analyzing cheeks for specular reflection inconsistencies.
</observe>
<hypothesize>
The cheeks shows abnormal specular-diffuse ratio. cheeks shows specular highlight inconsistency with soft top-right lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pa... | 1 | CONTRADICTION | 0.83 | [
120,
71,
463,
441
] | specular_reflection | 108 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01378 | deepfake | {"objects": [{"name": "eyes", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}... | Semantic Manifest:
Objects: eyes
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing eyes anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Eye region shows asymmetric feature limbal_ring inconsistency
</hypothesize>
<test>
The eyes shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — co... | <observe>
Analyzing eyes anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Eye region shows asymmetric feature limbal_ring inconsistency
</hypothesize>
<test>
The eyes shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — co... | 1 | CONTRADICTION | 0.86 | [
165,
120,
453,
494
] | anatomical | 264 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00124 | clean | {"objects": [{"name": "person", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera... | Semantic Manifest:
Objects: person
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | 1 | CONSISTENT | 0.91 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01335 | deepfake | {"objects": [{"name": "ears", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: ears
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Analyzing ears anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Ear shows asymmetric lobe — anatomical implausibility
</hypothesize>
<test>
The ears shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — confidence... | <observe>
Analyzing ears anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Ear shows asymmetric lobe — anatomical implausibility
</hypothesize>
<test>
The ears shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — confidence... | 1 | CONTRADICTION | 0.82 | [
199,
75,
427,
444
] | anatomical | 395 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01950 | deepfake | {"objects": [{"name": "face", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_h... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera low
Temporal: indoor, artificial_light | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 6: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confidenc... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 6: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confidenc... | 1 | CONTRADICTION | 0.9 | [
135,
111,
538,
329
] | composite | 536 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01045 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["impossible_angle"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "... | Semantic Manifest:
Objects: person, object
Lighting: top-left, shadows: ['impossible_angle'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing lighting: top-left source with shadows at ['impossible_angle'].
</observe>
<hypothesize>
This lighting_shadow pattern violates physical lighting laws. Single top-left source produces impossible shadow configuration: ['impossible_angle']
</hypothesize>
<test>
Light at top-left cannot produce shadows ... | <observe>
Analyzing lighting: top-left source with shadows at ['impossible_angle'].
</observe>
<hypothesize>
This lighting_shadow pattern violates physical lighting laws. Single top-left source produces impossible shadow configuration: ['impossible_angle']
</hypothesize>
<test>
Light at top-left cannot produce shadows ... | 1 | CONTRADICTION | 0.95 | [
74,
97,
435,
422
] | lighting_shadow | 58 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01482 | deepfake | {"objects": [{"name": "edge", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}... | Semantic Manifest:
Objects: edge
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 39 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.83
Regi... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 39 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.83
Regi... | 1 | CONTRADICTION | 0.83 | [
199,
61,
579,
318
] | frequency | 439 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00606 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.94 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00415 | clean | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01577 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspectiv... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera low
Temporal: indoor, daytime | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows perspective_violation for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [111, 117, 452, 358]
</ver... | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows perspective_violation for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [111, 117, 452, 358]
</ver... | 1 | CONTRADICTION | 0.85 | [
111,
117,
452,
358
] | pose_body | 465 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01978 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "low"}, "tempo... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera low
Temporal: indoor, natural_light | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 21: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 21: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | 1 | CONTRADICTION | 0.91 | [
79,
128,
406,
402
] | composite | 551 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00222 | clean | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspectiv... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | 1 | CONSISTENT | 0.95 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01848 | deepfake | {"objects": [{"name": "hands", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"},... | Semantic Manifest:
Objects: hands
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: m, o, t, i, o, n, _, b, l, u, r | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows optical_flow_discontinuity — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [9... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows optical_flow_discontinuity — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [9... | 1 | CONTRADICTION | 0.85 | [
93,
196,
413,
468
] | temporal_video | 525 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00955 | clean | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspectiv... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | 1 | CONSISTENT | 0.96 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00603 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.89 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01813 | deepfake | {"objects": [{"name": "face", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspective"... | Semantic Manifest:
Objects: face, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: c, o, m, p, r, e, s, s, i, o, n | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows lighting_flicker — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [77, 133, 54... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows lighting_flicker — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [77, 133, 54... | 1 | CONTRADICTION | 0.85 | [
77,
133,
542,
466
] | temporal_video | 522 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01615 | deepfake | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows pose_inconsistency for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [84, 188, 445, 451]
</verdict... | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows pose_inconsistency for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [84, 188, 445, 451]
</verdict... | 1 | CONTRADICTION | 0.87 | [
84,
188,
445,
451
] | pose_body | 485 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01575 | deepfake | {"objects": [{"name": "person", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "per... | Semantic Manifest:
Objects: person, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera low
Temporal: outdoor, daytime | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows proportion_distortion for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [140, 105, 546, 368]
</ver... | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows proportion_distortion for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [140, 105, 546, 368]
</ver... | 1 | CONTRADICTION | 0.85 | [
140,
105,
546,
368
] | pose_body | 480 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00451 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00339 | clean | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.94 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01470 | deepfake | {"objects": [{"name": "edge", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}... | Semantic Manifest:
Objects: edge
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 24 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.80
Regi... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 24 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.80
Regi... | 1 | CONTRADICTION | 0.8 | [
148,
115,
538,
485
] | frequency | 424 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00521 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.93 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01677 | deepfake | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": "outdoor_props"}, "perspective": {"vanishing_point": "center", "camera_height... | Semantic Manifest:
Objects: person
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=outdoor_props
Perspective: center vanishing point, camera eye-level
Temporal: i, n, d, o, o, r | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows impossible_background — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [58, 122, 552, ... | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows impossible_background — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [58, 122, 552, ... | 1 | CONTRADICTION | 0.86 | [
58,
122,
552,
304
] | environment | 502 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01629 | deepfake | {"objects": [{"name": "torso", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_... | Semantic Manifest:
Objects: torso
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows limb_elongation for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [132, 54, 438, 312]
</verdict> | <observe>
Analyzing scene for pose_body inconsistencies.
</observe>
<hypothesize>
Pattern violation: Body shows limb_elongation for camera height
</hypothesize>
<test>
Physical/logical inconsistency confirmed in pose_body domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [132, 54, 438, 312]
</verdict> | 1 | CONTRADICTION | 0.86 | [
132,
54,
438,
312
] | pose_body | 481 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01773 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": "wall"}, "perspe... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=wall
Perspective: center vanishing point, camera eye-level
Temporal: n, i, g, h, t, t, i, m, e | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows temporal_inconsistency — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.89
Region: [103, 156, 505... | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows temporal_inconsistency — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.89
Region: [103, 156, 505... | 1 | CONTRADICTION | 0.89 | [
103,
156,
505,
311
] | environment | 493 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00822 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "per... | Semantic Manifest:
Objects: person, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | 1 | CONSISTENT | 0.95 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01208 | deepfake | {"objects": [{"name": "skin", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "diffuse", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "per... | Semantic Manifest:
Objects: skin, background
Lighting: diffuse, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Analyzing skin for specular reflection inconsistencies.
</observe>
<hypothesize>
The skin shows abnormal specular-diffuse ratio. skin shows specular highlight inconsistency with harsh diffuse lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern d... | <observe>
Analyzing skin for specular reflection inconsistencies.
</observe>
<hypothesize>
The skin shows abnormal specular-diffuse ratio. skin shows specular highlight inconsistency with harsh diffuse lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern d... | 1 | CONTRADICTION | 0.85 | [
73,
142,
464,
321
] | specular_reflection | 110 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00288 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01771 | deepfake | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": "sky"}, "perspective": {"vanishing_point": "right", "camera_height": "... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=sky
Perspective: right vanishing point, camera high
Temporal: n, i, g, h, t, t, i, m, e | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.89
Region: [96, 109, 581, 374]
<... | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.89
Region: [96, 109, 581, 374]
<... | 1 | CONTRADICTION | 0.89 | [
96,
109,
581,
374
] | environment | 505 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01222 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Analyzing frequency domain for generation artifacts.
</observe>
<hypothesize>
GAN/diffusion fingerprint detected: Image shows latent trajectory embedding anomaly — LATTE fingerprint pattern
</hypothesize>
<test>
DCT analysis and spectral fingerprinting reveal AI generation traces. CIPHER/DiffusionPrint patter... | <observe>
Analyzing frequency domain for generation artifacts.
</observe>
<hypothesize>
GAN/diffusion fingerprint detected: Image shows latent trajectory embedding anomaly — LATTE fingerprint pattern
</hypothesize>
<test>
DCT analysis and spectral fingerprinting reveal AI generation traces. CIPHER/DiffusionPrint patter... | 1 | CONTRADICTION | 0.84 | [
122,
56,
568,
377
] | gan_fingerprint | 246 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00318 | clean | {"objects": [{"name": "person", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera... | Semantic Manifest:
Objects: person
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | 1 | CONSISTENT | 0.96 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01966 | deepfake | {"objects": [{"name": "face", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "pe... | Semantic Manifest:
Objects: face, background
Lighting: top-left, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera low
Temporal: outdoor, daytime | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 15: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 15: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | 1 | CONTRADICTION | 0.91 | [
123,
62,
437,
390
] | composite | 545 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01884 | deepfake | {"objects": [{"name": "hands", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"... | Semantic Manifest:
Objects: hands
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: m, o, t, i, o, n, _, b, l, u, r | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows optical_flow_discontinuity — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [1... | <observe>
Analyzing scene for temporal_video inconsistencies.
</observe>
<hypothesize>
Pattern violation: Frame shows optical_flow_discontinuity — temporal deepfake trace
</hypothesize>
<test>
Physical/logical inconsistency confirmed in temporal_video domain.
</test>
<verdict>
CONTRADICTION — confidence 0.85
Region: [1... | 1 | CONTRADICTION | 0.85 | [
195,
187,
491,
385
] | temporal_video | 525 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00269 | clean | {"objects": [{"name": "person", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera... | Semantic Manifest:
Objects: person
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01961 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "side", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspective": ... | Semantic Manifest:
Objects: person, object
Lighting: side, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 14: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 14: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | 1 | CONTRADICTION | 0.9 | [
55,
124,
448,
316
] | composite | 544 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00704 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.89 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00187 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.96 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00734 | clean | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspectiv... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | 1 | CONSISTENT | 0.92 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01164 | deepfake | {"objects": [{"name": "nose", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "side", "shadow_directions": ["bottom"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspective": {"vani... | Semantic Manifest:
Objects: nose, object
Lighting: side, shadows: ['bottom'], intensity: soft
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing nose for specular reflection inconsistencies.
</observe>
<hypothesize>
The nose shows abnormal specular-diffuse ratio. nose shows specular highlight inconsistency with soft side lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern detec... | <observe>
Analyzing nose for specular reflection inconsistencies.
</observe>
<hypothesize>
The nose shows abnormal specular-diffuse ratio. nose shows specular highlight inconsistency with soft side lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern detec... | 1 | CONTRADICTION | 0.85 | [
92,
112,
532,
344
] | specular_reflection | 106 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00797 | clean | {"objects": [{"name": "face", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-level"}, ... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | <observe>
Primary light source: top-left. Shadow direction: bottom-right. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-left light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-left → bottom-right. This follows Lambert's law. N... | 1 | CONSISTENT | 0.94 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01962 | deepfake | {"objects": [{"name": "face", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera_h... | Semantic Manifest:
Objects: face
Lighting: top-left, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera low
Temporal: indoor, artificial_light | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 0: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confidenc... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 0: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confidenc... | 1 | CONTRADICTION | 0.9 | [
183,
135,
505,
448
] | composite | 530 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00818 | clean | {"objects": [{"name": "person", "position": "center-left", "occludes": ["background"]}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "diffuse"}, "depth": {"foreground": ["person"], "midground": [], "background": ["wall"]}, "perspective": {"vanishing_point": "center", "camera... | Semantic Manifest:
Objects: person
Lighting: overhead, shadows: ['bottom'], intensity: diffuse
Depth: foreground=['person'], midground=[], background=['wall']
Perspective: center vanishing point, camera eye-level
Temporal: indoor, artificial_light | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: diffuse.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physica... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01206 | deepfake | {"objects": [{"name": "jawline", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-leve... | Semantic Manifest:
Objects: jawline
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['person'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing jawline for specular reflection inconsistencies.
</observe>
<hypothesize>
The jawline shows abnormal specular-diffuse ratio. jawline shows abnormal Retinex texture estimation for soft lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern... | <observe>
Analyzing jawline for specular reflection inconsistencies.
</observe>
<hypothesize>
The jawline shows abnormal specular-diffuse ratio. jawline shows abnormal Retinex texture estimation for soft lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SRI-Net pattern... | 1 | CONTRADICTION | 0.79 | [
81,
90,
512,
416
] | specular_reflection | 149 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00574 | clean | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspectiv... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | <observe>
Primary light source: overhead. Shadow direction: bottom. Intensity: overcast.
</observe>
<hypothesize>
The lighting configuration is physically consistent. overhead light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: overhead → bottom. This follows Lambert's law. No physic... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01173 | deepfake | {"objects": [{"name": "face_cheeks", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height"... | Semantic Manifest:
Objects: face_cheeks
Lighting: top-right, shadows: ['bottom-left'], intensity: harsh
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Analyzing face_cheeks for specular reflection inconsistencies.
</observe>
<hypothesize>
The face_cheeks shows abnormal specular-diffuse ratio. face_cheeks shows abnormal Retinex texture estimation for harsh lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SR... | <observe>
Analyzing face_cheeks for specular reflection inconsistencies.
</observe>
<hypothesize>
The face_cheeks shows abnormal specular-diffuse ratio. face_cheeks shows abnormal Retinex texture estimation for harsh lighting
</hypothesize>
<test>
Specular highlight geometry is inconsistent with stated light source. SR... | 1 | CONTRADICTION | 0.81 | [
94,
126,
557,
451
] | specular_reflection | 142 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01288 | deepfake | {"objects": [{"name": "skin", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "high"... | Semantic Manifest:
Objects: skin
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Analyzing frequency domain for generation artifacts.
</observe>
<hypothesize>
GAN/diffusion fingerprint detected: Image shows latent trajectory embedding anomaly — LATTE fingerprint pattern
</hypothesize>
<test>
DCT analysis and spectral fingerprinting reveal AI generation traces. CIPHER/DiffusionPrint patter... | <observe>
Analyzing frequency domain for generation artifacts.
</observe>
<hypothesize>
GAN/diffusion fingerprint detected: Image shows latent trajectory embedding anomaly — LATTE fingerprint pattern
</hypothesize>
<test>
DCT analysis and spectral fingerprinting reveal AI generation traces. CIPHER/DiffusionPrint patter... | 1 | CONTRADICTION | 0.81 | [
140,
94,
445,
360
] | gan_fingerprint | 243 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00873 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": ["background"]}, "per... | Semantic Manifest:
Objects: person, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=['background']
Perspective: left vanishing point, camera eye-level
Temporal: outdoor, daytime | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | <observe>
Primary light source: right. Shadow direction: left. Intensity: harsh.
</observe>
<hypothesize>
The lighting configuration is physically consistent. right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: right → left. This follows Lambert's law. No physical violations de... | 1 | CONSISTENT | 0.95 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01747 | deepfake | {"objects": [{"name": "person", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": "outdoor_props"}... | Semantic Manifest:
Objects: person, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=outdoor_props
Perspective: center vanishing point, camera eye-level
Temporal: w, i, n, t, e, r | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [81, 73, 429, 480]
</... | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.87
Region: [81, 73, 429, 480]
</... | 1 | CONTRADICTION | 0.87 | [
81,
73,
429,
480
] | environment | 495 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01433 | deepfake | {"objects": [{"name": "teeth", "position": "right", "occludes": []}, {"name": "object", "position": "left", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom"], "intensity": "overcast"}, "depth": {"foreground": ["person"], "midground": ["object"], "background": []}, "perspective... | Semantic Manifest:
Objects: teeth, object
Lighting: overhead, shadows: ['bottom'], intensity: overcast
Depth: foreground=['person'], midground=['object'], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, daytime | <observe>
Analyzing teeth anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Mouth region shows unnatural lip_texture pattern
</hypothesize>
<test>
The teeth shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — confidence 0.... | <observe>
Analyzing teeth anatomical structure.
</observe>
<hypothesize>
Anatomical violation detected: Mouth region shows unnatural lip_texture pattern
</hypothesize>
<test>
The teeth shows characteristics typical of AI-generated content. Distortion-5K pattern confirmed.
</test>
<verdict>
CONTRADICTION — confidence 0.... | 1 | CONTRADICTION | 0.86 | [
136,
77,
593,
406
] | anatomical | 298 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01726 | deepfake | {"objects": [{"name": "person", "position": "center", "occludes": []}, {"name": "background", "position": "background", "occludes": []}], "lighting": {"primary_source": "right", "shadow_directions": ["left"], "intensity": "harsh"}, "depth": {"foreground": ["person"], "midground": [], "background": "ceiling"}, "perspect... | Semantic Manifest:
Objects: person, background
Lighting: right, shadows: ['left'], intensity: harsh
Depth: foreground=['person'], midground=[], background=ceiling
Perspective: left vanishing point, camera eye-level
Temporal: i, n, d, o, o, r | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [155, 155, 477, 413]
... | <observe>
Analyzing scene for environment inconsistencies.
</observe>
<hypothesize>
Pattern violation: Scene shows season_mismatch — context impossibility
</hypothesize>
<test>
Physical/logical inconsistency confirmed in environment domain.
</test>
<verdict>
CONTRADICTION — confidence 0.86
Region: [155, 155, 477, 413]
... | 1 | CONTRADICTION | 0.86 | [
155,
155,
477,
413
] | environment | 490 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01992 | deepfake | {"objects": [{"name": "hands", "position": "center", "occludes": []}], "lighting": {"primary_source": "overhead", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "high"}, "tem... | Semantic Manifest:
Objects: hands
Lighting: overhead, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera high
Temporal: indoor, natural_light | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 22: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | <observe>
Analyzing scene for composite inconsistencies.
</observe>
<hypothesize>
Pattern violation: Composite pattern 22: Multiple physical violations across lighting, anatomy, and perspective
</hypothesize>
<test>
Physical/logical inconsistency confirmed in composite domain.
</test>
<verdict>
CONTRADICTION — confiden... | 1 | CONTRADICTION | 0.9 | [
85,
101,
447,
397
] | composite | 552 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00176 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.92 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
clean_00102 | clean | {"objects": [{"name": "person", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-right", "shadow_directions": ["bottom-left"], "intensity": "soft"}, "depth": {"foreground": ["person"], "midground": ["floor"], "background": []}, "perspective": {"vanishing_point": "right", "camera_height": "hig... | Semantic Manifest:
Objects: person
Lighting: top-right, shadows: ['bottom-left'], intensity: soft
Depth: foreground=['person'], midground=['floor'], background=[]
Perspective: right vanishing point, camera high
Temporal: outdoor, evening | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | <observe>
Primary light source: top-right. Shadow direction: bottom-left. Intensity: soft.
</observe>
<hypothesize>
The lighting configuration is physically consistent. top-right light casts shadows in the expected direction.
</hypothesize>
<test>
Verifying physics: top-right → bottom-left. This follows Lambert's law. ... | 1 | CONSISTENT | 0.9 | [
100,
100,
500,
400
] | none | 0 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
fake_01493 | deepfake | {"objects": [{"name": "background", "position": "center", "occludes": []}], "lighting": {"primary_source": "top-left", "shadow_directions": ["bottom-right"], "intensity": "soft"}, "depth": {"foreground": ["face"], "midground": [], "background": []}, "perspective": {"vanishing_point": "center", "camera_height": "eye-lev... | Semantic Manifest:
Objects: background
Lighting: top-left, shadows: ['bottom-right'], intensity: soft
Depth: foreground=['face'], midground=[], background=[]
Perspective: center vanishing point, camera eye-level
Temporal: indoor, natural_light | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 3 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.83
Regio... | <observe>
Analyzing scene for frequency inconsistencies.
</observe>
<hypothesize>
Pattern violation: Region shows DCT coefficient anomaly pattern 3 — spectral fingerprint violation
</hypothesize>
<test>
Physical/logical inconsistency confirmed in frequency domain.
</test>
<verdict>
CONTRADICTION — confidence 0.83
Regio... | 1 | CONTRADICTION | 0.83 | [
168,
190,
418,
374
] | frequency | 403 | You are the Inquisitor — an adversarial deepfake detection agent.
Analyze scene manifests for physical inconsistencies using 500+ modern deepfake patterns.
PATTERN CATEGORIES:
- Lighting & Shadow (1-100): Shadow direction violations, light source impossibilities
- Specular Reflection (101-170): Specular-diffuse ratio ... |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- 100