id stringlengths 36 72 | question_id stringlengths 8 30 | sample_id stringlengths 12 30 | path_id stringlengths 21 21 ⌀ | system stringclasses 5
values | conversations listlengths 2 102 | images listlengths 0 12 |
|---|---|---|---|---|---|---|
q_000007_sample_path_818f90c3fe7834e5_path_818f90c3fe7834e5_971fb5dad7 | q_000007 | sample_path_818f90c3fe7834e5 | path_818f90c3fe7834e5 |
You are a multimodal research agent. Solve the user request through explicit step-by-step reasoning and tool use.
Requirements:
1. Every factual claim in the visible solution must be supported by the question, the image, or tool-returned evidence.
2. If the evidence is insufficient, state the uncertainty and continue... | [
{
"from": "human",
"value": "In a 14th-century full-page manuscript miniature depicting the enthroned emperor whose death in the late 12th century brought renewed prestige to an alliance of cities, how many eagles are on the shield in the upper left?"
},
{
"from": "gpt",
"value": "<thinking>\nTh... | [
"images/b1/b170ce20548e3bf43f1d1a53b9a7ada0cb954f964cc34aa526809bbfb5993ee3.jpg",
"images/f5/f516e8766c6414f3f53a1f3de4e04e6cbe129800c17d643982800428225d766c.jpg",
"images/7c/7c929d6453b8a65d6a2d78d6407438f70c2328f56cccc46a01e44cfb01b0dbf6.jpg"
] |
q_000013_sample_path_b50a471615ed01d9_path_b50a471615ed01d9_02c59a1803 | q_000013 | sample_path_b50a471615ed01d9 | path_b50a471615ed01d9 |
You are a multimodal research agent. Solve the user request through explicit step-by-step reasoning and tool use.
Requirements:
1. Every factual claim in the visible solution must be supported by the question, the image, or tool-returned evidence.
2. If the evidence is insufficient, state the uncertainty and continue... | [
{
"from": "human",
"value": "In this frontal view of an ironclad gunboat displayed under a canopy, how many cannons can be seen pointing forward from the ship's main armored casemate?\n<image>"
},
{
"from": "gpt",
"value": "<thinking>\nThe image is clear enough to answer directly from visual ins... | [
"images/b2/b2964af85fd810db4337f1d168c4e957c2c306223c70eb322fbbcf70243ac603.png"
] |
q_000012_sample_path_e97c01e3bc473e84_path_e97c01e3bc473e84_a6632ef48b | q_000012 | sample_path_e97c01e3bc473e84 | path_e97c01e3bc473e84 |
You are a multimodal research agent. Solve the user request through explicit step-by-step reasoning and tool use.
Requirements:
1. Every factual claim in the visible solution must be supported by the question, the image, or tool-returned evidence.
2. If the evidence is insufficient, state the uncertainty and continue... | [
{
"from": "human",
"value": "Consider the driver on the left in the image, who has 'MAX' written on his race suit. During the podium celebration for his first Formula One victory, a sponsor's name was visible on the screen behind him. In 1974, upon a request from a rally racing team, for which specific car ... | [
"images/3a/3aeb7cb03def9f16025787d5ed70da989500efd5f791b8c0d37d49f4f0b657cb.png",
"images/3d/3dcd61f68bb58f6f8fc577be89d12b950f4afacb2e24b2eed68ccbf59f7f85a2.jpg",
"images/96/964e00e0c9492c4c39d31687d2c63920a383e08acaf58d5891fe0df5fe6ab018.jpg",
"images/df/df3284c11cadbc673d86b07e8a3d5cd2183f74f6605fa94229bbf... |
q_000005_sample_path_ab414ea811902a38_path_ab414ea811902a38_8dadb75d3b | q_000005 | sample_path_ab414ea811902a38 | path_ab414ea811902a38 |
You are a multimodal research agent. Solve the user request through explicit step-by-step reasoning and tool use.
Requirements:
1. Every factual claim in the visible solution must be supported by the question, the image, or tool-returned evidence.
2. If the evidence is insufficient, state the uncertainty and continue... | [
{
"from": "human",
"value": "During a speech given by a head of state to a legislative body, in response to the military action sanctioned by the shorter man in the formal suit seen in the image standing next to Douglas MacArthur, what object sits on the desk directly in front of the bald man seated in the ... | [
"images/6f/6f891752c62d83709252a9daa00abf2dda6605afa6b44060f9a1216c63bf0eb0.png",
"images/40/40ce2c8d83dc4826d64ae75e37f7a5b8183a4f90fbea5782b08618976dcedf7e.jpg"
] |
q_000003_sample_path_2c53268c324d3c73_path_2c53268c324d3c73_fe957370a9 | q_000003 | sample_path_2c53268c324d3c73 | path_2c53268c324d3c73 |
You are a multimodal research agent. Solve the user request through explicit step-by-step reasoning and tool use.
Requirements:
1. Every factual claim in the visible solution must be supported by the question, the image, or tool-returned evidence.
2. If the evidence is insufficient, state the uncertainty and continue... | [
{
"from": "human",
"value": "The artist whose signature is on the horse's breastplate strap in the provided image painted a famous fresco in the Apostolic Palace. In that fresco, what kind of footwear is worn by the brooding figure depicted writing while leaning on a marble block in the foreground?\n<image>... | [
"images/f7/f7cf2c32e15f601f418092e21b2fbcdf173d10e365551a0f896ac6295d31d6f8.png",
"images/6b/6b451fd4532066291e9e7aa203faa6fd8414be74518b70906ba40d38fe28a74e.jpg",
"images/e4/e46ae5976112d2a7227f713963e4115b2caacb982bcbe4905126a939edf10a36.jpg"
] |
q_000009_sample_path_ad74932a4e2be978_path_ad74932a4e2be978_fb27054120 | q_000009 | sample_path_ad74932a4e2be978 | path_ad74932a4e2be978 | "\nYou are a multimodal research agent. Solve the user request through explicit step-by-step reasoni(...TRUNCATED) | [{"from":"human","value":"In the capital of the country whose national railway operates the wrecked (...TRUNCATED) | ["images/56/568285b5e763ea2bed547cc389c64195c4e11b4c0e3b63bde0a4a498b5a72a7e.png","images/6a/6ac86b5(...TRUNCATED) |
q_000004_sample_path_429917eca42fd884_path_429917eca42fd884_a3dc7c4817 | q_000004 | sample_path_429917eca42fd884 | path_429917eca42fd884 | "\nYou are a multimodal research agent. Solve the user request through explicit step-by-step reasoni(...TRUNCATED) | [{"from":"human","value":"According to a recent study from a major research center, what percentage (...TRUNCATED) | ["images/c2/c2c9d59e84347f798ebdf21fcf76083dffa1234bf4677102d118ad867e7ff53f.png","images/94/94867dc(...TRUNCATED) |
q_000024_sample_path_7e12019c6bbc91dc_path_7e12019c6bbc91dc_df21ade26a | q_000024 | sample_path_7e12019c6bbc91dc | path_7e12019c6bbc91dc | "\nYou are a multimodal research agent. Solve the user request through explicit step-by-step reasoni(...TRUNCATED) | [{"from":"human","value":"At the museum for the brand identified on the oval sign in this image, wha(...TRUNCATED) | ["images/26/261587310c7ae4c184f881a088200bba1e1818bfbefcb255b79754401d98e769.png","images/a7/a74a9c8(...TRUNCATED) |
q_000008_sample_path_3a9d6c8e2b791808_path_3a9d6c8e2b791808_b88ce70076 | q_000008 | sample_path_3a9d6c8e2b791808 | path_3a9d6c8e2b791808 | "\nYou are a multimodal research agent. Solve the user request through explicit step-by-step reasoni(...TRUNCATED) | [{"from":"human","value":"While still an active player, a teammate of the captain of the winning tea(...TRUNCATED) | ["images/b9/b98072c2297e876fa49e65c2774530ba65fb35897909a886022cbd4497ebaeaa.png","images/90/9003568(...TRUNCATED) |
q_000030_sample_path_6f8ba7ccfa513cc5_path_6f8ba7ccfa513cc5_3d210970ec | q_000030 | sample_path_6f8ba7ccfa513cc5 | path_6f8ba7ccfa513cc5 | "\nYou are a multimodal research agent. Solve the user request through explicit step-by-step reasoni(...TRUNCATED) | [{"from":"human","value":"A U.S. post office building was dedicated in a certain city in the early 1(...TRUNCATED) | [
"images/11/111dfeda023ca5424acba2aa22fe33571f2d074de68270135a11c04f4c8769c1.jpg"
] |
End of preview. Expand in Data Studio
Jev-SFT
Public dataset backup: 16004 rows; 33435 external image files.
Conversations, answers, row order, IDs and RL extra_info are preserved. SFT/RL image paths are relocated; image bytes are archived unchanged. Image files are hashed by original path for naming, not deduplicated by image content.
Download and restore
hf download hansQAQ/Jev-SFT --repo-type dataset --local-dir ./Jev-SFT
python ./Jev-SFT/restore_jev_dataset.py --dataset-dir ./Jev-SFT
Point LLaMA-Factory dataset_dir and media_dir to this downloaded directory; dataset_info.json is included.
Keep metadata/: it records the original SFT merge and alignment exclusions. The original 16004 SFT rows are preserved; the training cache's 8-row exclusion is documented separately.
Full 24000 candidate conversations
The original training dataset remains available as jev_sft_16004. jev_sft_24000 is the separate full candidate version, including records excluded from final training. Both use the same images directory; extra image archives contain only missing images.
hf download hansQAQ/Jev-SFT --repo-type dataset --local-dir ./Jev-SFT
python ./Jev-SFT/restore_jev_sft_full24000.py --dataset-dir ./Jev-SFT
LLaMA-Factory configuration after restoration:
dataset_dir: /absolute/path/to/Jev-SFT
media_dir: /absolute/path/to/Jev-SFT
dataset: jev_sft_24000
Use dataset: jev_sft_16004 for the original dataset. All image paths are already relative to the repository root; no manual merging or path rewriting is needed.
- Downloads last month
- 75