Dataset Viewer
Auto-converted to Parquet Duplicate
pid
string
question
string
image
string
decoded_image
image
choices
sequence
unit
string
precision
float64
answer
string
question_type
string
answer_type
string
metadata
dict
query
string
subquestions
string
86
Which model has the highest Acc score when Pretrain Loss is equal to 1.80?
images/86.jpg
[ "ICL", "SFT", "SFT 1/8", "RFT k=100", "RFT k=25", "RET k=6", "RFT U13B" ]
null
null
RFT U13B
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 1046, "img_width": 1734, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model has the highest Acc score when Pretrain Loss is equal to 1.80? Choices: (A) ICL (B) SFT (C) SFT 1/8 (D) RFT k=100 (E) RFT k=25 (F) RET k=6 (G) RFT U13B
```json [ { "question": "What Pretrain Loss is marked by the star?", "answer": "1.80" }, { "question": "What model is represented by the cyan line?", "answer": "RFT U13B" }, { "question": "At 1.80 Pretrain Loss, what's RFT U13B Acc?", "answer": "50" }, { "question": "What model...
188
How many methods in the table achieve an A-847 score higher than 20.0?
images/188.jpg
null
null
null
3
free_form
integer
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 634, "img_width": 2226, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: How many methods in the table achieve an A-847 score higher than 20.0?
```json [ { "question": "What is the column labeled 'A-847'?", "answer": "grounding mIoU" }, { "question": "What A-847 score does 'ALIGN [38, 28]' achieve?", "answer": "17.8" }, { "question": "What A-847 score does 'LSeg+ [46, 28]' achieve?", "answer": "10.5" }, { "question": "...
197
How many sequences have negative Influence Scores?
images/197.jpg
null
null
null
2
free_form
integer
{ "category": "math-targeted-vqa", "context": "bar chart", "grade": "college", "img_height": 772, "img_width": 1766, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: How many sequences have negative Influence Scores?
```json [ { "question": "What is on the x-axis?", "answer": "Sequences" }, { "question": "What is on the y-axis?", "answer": "Influence Scores" }, { "question": "What does the orange bar represent?", "answer": "Mandarin Followed" }, { "question": "How many Mandarin followed by ...
231
Which model has the overall best ImageNet 10shot Accuracy score across different training steps?
images/231.jpg
[ "Identity", "Uniform", "Uniform / Soft", "Soft / Uniform", "Soft", "Dense" ]
null
null
Soft
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 988, "img_width": 2002, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model has the overall best ImageNet 10shot Accuracy score across different training steps? Choices: (A) Identity (B) Uniform (C) Uniform / Soft (D) Soft / Uniform (E) Soft (F) Dense
```json [ { "question": "What metric does the right-side graph represent?", "answer": "10shot Accuracy" }, { "question": "What is the highest ImageNet 10shot Accuracy achieved?", "answer": "About 0.74" }, { "question": "What color represents the 'Soft' model?", "answer": "Purple" }, ...
241
How many models in the table have a model size larger than 10B?
images/241.jpg
null
null
null
11
free_form
integer
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 1184, "img_width": 1570, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: How many models in the table have a model size larger than 10B?
```json [ { "question": "What is the model size of 'wizardlm-30b'?", "answer": "30B" }, { "question": "What is the model size of 'wizardcoder-15b'?", "answer": "15B" }, { "question": "What is the model size of 'wizardlm-13b'?", "answer": "13B" }, { "question": "What is the mode...
278
Among the following objects, which one has the best PSNR score?
images/278.jpg
[ "Lego", "Mats", "Mic", "Ship" ]
null
null
Mic
multi_choice
text
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 940, "img_width": 1478, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Among the following objects, which one has the best PSNR score? Choices: (A) Lego (B) Mats (C) Mic (D) Ship
```json [ { "question": "What is the highest PSNR score listed?", "answer": "33.4" }, { "question": "Which object achieves a PSNR score of 33.2 using DMTet?", "answer": "Hotdog" }, { "question": "Which method yields a higher PSNR score for the 'Ship' object?", "answer": "DMTet" }, ...
288
In which of the following value ranges of λ2 does the percentage of Attack Effectiveness begin to be lower than that of Diversity?
images/288.jpg
[ "0.0 - 0.2", "0.2 - 0.4", "0.4 - 0.6", "0.6 - 0.8", "0.8 - 1.0" ]
null
null
0.0 - 0.2
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 606, "img_width": 2144, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: In which of the following value ranges of λ2 does the percentage of Attack Effectiveness begin to be lower than that of Diversity? Choices: (A) 0.0 - 0.2 (B) 0.2 - 0.4 (C) 0.4 - 0.6 (D) 0.6 - 0.8 (E) 0.8 - 1....
```json [ { "question": "At λ2=0.0 for Stable Diffusion, what is Attack Effectiveness percentage?", "answer": "Around 85" }, { "question": "At λ2=0.0 for Stable Diffusion, what is the Diversity percentage?", "answer": "Around 50" }, { "question": "For Strong Safe SD, when is Attack Effecti...
301
Which model performs the best overall across the three stages in terms of Messenger training performance?
images/301.jpg
[ "Dynalang", "EMMA", "R2D2", "IMPALA" ]
null
null
Dynalang
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 524, "img_width": 2012, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model performs the best overall across the three stages in terms of Messenger training performance? Choices: (A) Dynalang (B) EMMA (C) R2D2 (D) IMPALA
```json [ { "question": "What is the maximum score achieved by Dynalang in Stage 1?", "answer": "1.0" }, { "question": "In Stage 3, which model has the lowest score?", "answer": "R2D2" }, { "question": "Which model shows the most consistent high performance across all stages?", "answer...
340
Which of the following domains has the most number of BPE Tokens?
images/340.jpg
[ "Legal ", "Code ", "Conversational ", "Math ", "Science", "Books ", "News ", "Encyclopedic" ]
null
null
Science
multi_choice
text
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 1176, "img_width": 2142, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which of the following domains has the most number of BPE Tokens? Choices: (A) Legal (B) Code (C) Conversational (D) Math (E) Science (F) Books (G) News (H) Encyclopedic
```json [ { "question": "What domain is associated with GitHub?", "answer": "Code" }, { "question": "What is the BPE Tokens value for Encyclopedic?", "answer": "37.0" }, { "question": "What is the BPE token value for the domain Legal with public domain license?", "answer": "27.1" }, ...
349
Which capability boasts the highest proportion (%)?
images/349.jpg
[ "Rec", "OCR", "Know", "Gen", "Spat", "Math" ]
null
null
Rec
multi_choice
text
{ "category": "math-targeted-vqa", "context": "bar chart", "grade": "college", "img_height": 1348, "img_width": 1704, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which capability boasts the highest proportion (%)? Choices: (A) Rec (B) OCR (C) Know (D) Gen (E) Spat (F) Math
```json [ { "question": "What type of capabilities does (a) represent?", "answer": "Each capability" }, { "question": "What is the proportion for Recognition?", "answer": "68.8" }, { "question": "What is the proportion for OCR?", "answer": "44.0" }, { "question": "What is the p...
394
Which of the following environments has the least GPU days for training?
images/394.jpg
[ "HomeGrid", "Msgr S1", "Msgr S2", "Msgr S3", "VLN", "LangRoom" ]
null
null
LangRoom
multi_choice
text
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 858, "img_width": 1854, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which of the following environments has the least GPU days for training? Choices: (A) HomeGrid (B) Msgr S1 (C) Msgr S2 (D) Msgr S3 (E) VLN (F) LangRoom
```json [ { "question": "What metric measures training time?", "answer": "GPU days" }, { "question": "What is HomeGrid's training time?", "answer": "3.75" }, { "question": "What is VLN's training time?", "answer": "16" }, { "question": "What is LangRoom's training time?", "...
400
At which Episode ID does the Retroformer attain its peak Success rate (%)?
images/400.jpg
[ "1.0", "1.5", "2.0", "2.5", "3.0", "3.5", "4.0" ]
null
null
4.0
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 942, "img_width": 1196, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: At which Episode ID does the Retroformer attain its peak Success rate (%)? Choices: (A) 1.0 (B) 1.5 (C) 2.0 (D) 2.5 (E) 3.0 (F) 3.5 (G) 4.0
```json [ { "question": "What is plotted on the x-axis?", "answer": "Episode ID" }, { "question": "What is the highest Episode ID shown?", "answer": "4.0" }, { "question": "What is plotted on the y-axis?", "answer": "Success rate" }, { "question": "What is the success rate at E...
470
Which of the following models has the lowest KS Rollout Loss overall?
images/470.jpg
[ "Baseline", "Diffusion", "PDE-Refiner", "Pushforward" ]
null
null
PDE-Refiner
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 854, "img_width": 1422, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which of the following models has the lowest KS Rollout Loss overall? Choices: (A) Baseline (B) Diffusion (C) PDE-Refiner (D) Pushforward
```json [ { "question": "Which model is represented by the purple line?", "answer": "PDE-Refiner" }, { "question": "Which model has the lowest loss at rollout time step 0?", "answer": "PDE-Refiner" }, { "question": "What is the KS Rollout Loss's y-axis label?", "answer": "MSE Loss" }...
542
How many models in the figure achieve an Acc score greater than 60?
images/542.jpg
null
null
null
2
free_form
integer
{ "category": "math-targeted-vqa", "context": "scatter plot", "grade": "college", "img_height": 1358, "img_width": 1690, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: How many models in the figure achieve an Acc score greater than 60?
```json [ { "question": "What is the Acc score of GPT-4?", "answer": "90" }, { "question": "What is the Acc score of LLaMA2-70B?", "answer": "60" }, { "question": "What is the Acc score of LLaMA-65B?", "answer": "55" }, { "question": "What is the Acc score of LLaMA2-13B?", ...
552
How many queries have a p-value lower than 0.50?
images/552.jpg
null
null
null
4
free_form
integer
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 330, "img_width": 1726, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: How many queries have a p-value lower than 0.50?
```json [ { "question": "What is the p-value for 'objective'?", "answer": "0.01" }, { "question": "What is the p-value for 'paperclips'?", "answer": "0.43" }, { "question": "What is the p-value for 'bullet'?", "answer": "0.69" }, { "question": "What is the p-value for 'water'?"...
565
Which model can achieve the best ImageNet 10-shot Accuracy score?
images/565.jpg
[ "Soft MoE", "Experts Choice", "Tokens Choice", "Dense" ]
null
null
Soft MoE
multi_choice
text
{ "category": "math-targeted-vqa", "context": "line plot", "grade": "college", "img_height": 978, "img_width": 1966, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model can achieve the best ImageNet 10-shot Accuracy score? Choices: (A) Soft MoE (B) Experts Choice (C) Tokens Choice (D) Dense
```json [ { "question": "Which subplot shows ImageNet accuracy?", "answer": "subplot (b)" }, { "question": "What model has highest accuracy?", "answer": "Soft MoE" }, { "question": "What is the accuracy of Soft MoE?", "answer": "above 0.8" }, { "question": "What has lowest Imag...
793
What is the highest Elo score for the agent using an offline RL algorithm?
images/793.jpg
null
null
null
1578
free_form
integer
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 1056, "img_width": 1922, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end. Question: What is the highest Elo score for the agent using an offline RL algorithm?
```json [ { "question": "Which agents use Offline RL?", "answer": "Pale blue" }, { "question": "What is MuZero's Elo score?", "answer": "1578" }, { "question": "Elo of Offline Actor-Critic?", "answer": "1548" }, { "question": "What is AlphaStar Final's Elo?", "answer": "296...
895
Which model has the lowest Audio-Audio Similarity and Text-Audio Similarity scores overall?
images/895.jpg
[ "MusicLDM (mix-up)", "MusicLDM (original)", "MusicLDM (BLM)", "MusicLDM (BAM)", "MuBERT" ]
null
null
MuBERT
multi_choice
text
{ "category": "math-targeted-vqa", "context": "violin plot", "grade": "college", "img_height": 682, "img_width": 1882, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model has the lowest Audio-Audio Similarity and Text-Audio Similarity scores overall? Choices: (A) MusicLDM (mix-up) (B) MusicLDM (original) (C) MusicLDM (BLM) (D) MusicLDM (BAM) (E) MuBERT
```json [ { "question": "What is the lowest similarity score for 'Audio-Audio' according to the violin plots?", "answer": "Around 0.0" }, { "question": "Which model has the lowest median for Audio-Audio Similarity?", "answer": "MuBERT" }, { "question": "Which model has the lowest median fo...
923
Which model achieves the highest score in terms of Rec?
images/923.jpg
[ "Transformers Agent (GPT-4)", "LLaMA-Adapter v2-7B", "LLaVA-7B", "Otter-9B ", "MM-ReAct-GPT-3.5", "LLaVA-13B (LLaMA-2)", "MM-ReAct-GPT-4" ]
null
null
LLaVA-13B (LLaMA-2)
multi_choice
text
{ "category": "math-targeted-vqa", "context": "table", "grade": "college", "img_height": 1056, "img_width": 1910, "language": "english", "skills": [ "scientific reasoning", "statistical reasoning" ], "source": "PaperQA", "split": "testmini", "task": "figure question answering" }
Hint: Please answer the question and provide the correct option letter, e.g., A, B, C, D, at the end. Question: Which model achieves the highest score in terms of Rec? Choices: (A) Transformers Agent (GPT-4) (B) LLaMA-Adapter v2-7B (C) LLaVA-7B (D) Otter-9B (E) MM-ReAct-GPT-3.5 (F) LLaVA-13B (LLaMA-2) (G) MM-ReAct-GPT...
```json [ { "question": "What metric does 'Rec' stand for?", "answer": "Recall" }, { "question": "What is the highest Rec score shown?", "answer": "39.2" }, { "question": "Which model achieved a 'Rec' score of 39.2?", "answer": "LLaVA-13B" }, { "question": "What is the second h...
README.md exists but content is empty.
Downloads last month
1