Dataset Viewer
Auto-converted to Parquet Duplicate
data_source
stringclasses
1 value
prompt
listlengths
2
2
⌀
images
listlengths
2
10
ability
stringclasses
1 value
reward_model
dict
extra_info
dict
teacher_prompt
listlengths
2
2
⌀
teacher_images
listlengths
2
10
⌀
teacher_negative_prompt
listlengths
2
2
⌀
teacher_negative_images
listlengths
2
10
⌀
ve_verified_evidence
bool
2 classes
ve_evidence
stringlengths
0
555
⌀
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12370.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12371.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12372.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "be careful with him , folks . that fall didn ' t do dirty any good", "oh dirty will will be o . k ., boss what s left of him .", "wer wor th -?", "what ' s gone" ], "family": "visual_seq...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12370.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12371.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12372.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12370.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12371.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/12372.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character is depicted with a dazed expression and a speech bubble containing the text "WER WOR TH - ?" </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3536.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3537.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 5911, "num_images": 4, "original_answer": "True", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "M4-...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3536.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3537.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3536.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3537.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | Both images depict an interior space with white walls and open architectural elements. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14425.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14426.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14427.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "it is best that you mar him , mistress poverty is a harsh task . master !", "you came to marry a young man ... but i flatter myself i still have the vigor of youth !", "she ' s beautiful ! hmm",...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14425.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14426.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14427.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14425.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14426.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/14427.jpg" }, { "path": "../../d...
true
<evidence> E1 | A woman in a red dress is shown looking up at a man, with a speech bubble containing the text "she ' s beautiful ! hmm" </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1904.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1905.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "property_consistency_choice", "index": 4502, "num_images": 4, "original_answer": "True", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1904.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1905.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1904.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1905.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
true
<evidence> E1 | All four images display black, rectangular electronic devices with control interfaces. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4482.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4483.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "property_consistency_choice", "index": 1824, "num_images": 2, "original_answer": "True", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4482.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4483.jpg" } ]
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4482.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4483.jpg" } ]
true
<evidence> E1 | Both images display a mechanical gear with teeth surrounding a central circular hole. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img246.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img249.jpg" }, { "path": "../../data/images/man...
multi_image_cot
{ "ground_truth": "F", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "F", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 1228, "num_images": 10, "original_answer": "Answer: Image 6", "p...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img246.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img249.jpg" }, { "path": "../../data/images/man...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img246.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_NWvTOHoShYk-shot10_3/img249.jpg" }, { "path": "../../data/images/man...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1580.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1581.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "property_consistency_choice", "index": 9253, "num_images": 4, "original_answer": "True", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1580.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1581.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1580.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/1581.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13140.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13141.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13142.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "re a pretty smart little cookie . aren ' t you ?", "honey ? and you ' ll still let me take you to the movies tomorrow night ?", "as a matter of fact i didn ' t know i don ' t expect you to belie...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13140.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13141.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13142.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13140.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13141.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/13142.jpg" }, { "path": "../../d...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5322.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5323.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "314002840", "314002839", "314002837", "314002838" ], "family": "document_ocr_choice", "index": 8318, "num_images": 2, "original_answer": "314002838", "p1_version": "plain_objecti...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5322.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5323.jpg" } ]
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5322.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5323.jpg" } ]
true
<evidence> E1 | The number 314002838 is printed vertically along the left margin of both images. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19170.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19171.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19172.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "greetings mr monk .", "sure look glamorous today", "we gotta run it ' s a super picture", "gallant mr monk" ], "family": "visual_sequence_choice", "index": 5167, "num_images": 5, "...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19170.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19171.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19172.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19170.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19171.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19172.jpg" }, { "path": "../../d...
true
<evidence> E1 | A hippo character is walking on the sidewalk while wearing a pink dress and carrying a green handbag. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15265.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15266.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15267.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "the flying coffin closes the gap between the two .", "it ' s time to call a halt to this wild goose chase !", "ho ! so the monkeys grabbed his tail and run ? h mm this next job should be easier ...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15265.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15266.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15267.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15265.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15266.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/15267.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character in the cockpit is flying at high speed and is responding to an enemy pilot who has just been stopped. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19565.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19566.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19567.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "aim for the tires , men too late ... the car ' s speeding away . quick ... into your cars they won ' t get far !", "recklessly , the car that carried floyd and bird well kept moving ahead then", ...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19565.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19566.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19567.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19565.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19566.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19567.jpg" }, { "path": "../../d...
true
<evidence> E1 | A police officer is aiming a rifle at a yellow car while shouting "Bang, Bam!" in the air. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img54.jpg" }, { "path": "../../data/images/mantis/...
multi_image_cot
{ "ground_truth": "H", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "H", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 5464, "num_images": 10, "original_answer": "Answer: H", "p1_vers...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img54.jpg" }, { "path": "../../data/images/mantis/...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6imbMZ2t7Qc-shot21_1/img54.jpg" }, { "path": "../../data/images/mantis/...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1606.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1607.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "Pre-registration fee for ASBT meeting in Phoenix for Larry McGhee", "Preliminary registration fee for ASMBT meeting in Phoenix for Larry McGhee", "pre-registration fee for ASMBT meeting in Phoen...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1606.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1607.jpg" } ]
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1606.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1607.jpg" } ]
true
<evidence> E1 | The description column reads "Pre-registration fee for assbt meeting in Phoenix for larry McGhee" </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23755.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23756.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23757.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "bah you are welcome to your chip of stone hag have tried to it is worthless", "i do not think ach , now i know what it 15 why did i not think of it before", "tali on van from the ca begins to gl...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23755.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23756.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23757.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23755.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23756.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23757.jpg" }, { "path": "../../d...
true
<evidence> E1 | The masked speech bubble in the final panel is located between two characters who appear to be reacting to a discovery. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6337.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "you you poor man ! you are hurt , let me help you .", "get away , you demons a", "you must be an angel", "you are very , very kind" ], "family": "visual_sequence_choice", "index": 1151...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6337.jpg" }, { "path": "../../data...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6337.jpg" }, { "path": "../../data...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8985.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8986.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8987.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "w up ! outa gas", "maybe we better try to borrow some at this farmhouse", "my land boy ! what ' s that on your face", "oboy this is a swell lookin ' town to operate our medicine show" ], ...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8985.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8986.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8987.jpg" }, { "path": "../../data...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8985.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8986.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8987.jpg" }, { "path": "../../data...
true
<evidence> E1 | The character states their medicine show must close and a sound effect "wup! outa gas" is depicted. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 464, "num_images": 10, "original_answer": "Answer: Image 4", "p1...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2304-shot4_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
true
<evidence> E1 | The woman's hand is holding the edge of a cookie that is resting on the white rectangular plate. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16325.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16326.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16327.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "there it is , tom -- the birth mark on his shoulder !", "in this world ... but ..", "eddie -- no dont go through that screen !", "time max hall and i owned a little hat store on high street ...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16325.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16326.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16327.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16325.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16326.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16327.jpg" }, { "path": "../../d...
false
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img26.jpg" }, { "path": "../../data/images/mantis...
multi_image_cot
{ "ground_truth": "G", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "G", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 10882, "num_images": 10, "original_answer": "Answer: G", "p1_ver...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img26.jpg" }, { "path": "../../data/images/mantis...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_LKYioW0-ytk-shot80_0/img26.jpg" }, { "path": "../../data/images/mantis...
true
<evidence> E1 | The toddler in the blue shirt is sitting on the floor with both hands empty, holding them in front of him while looking into the gift bag. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25955.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25956.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25957.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "ver but the memory", "halt . identify yourself .", "even sounds like an order", "and tired of hiding beam / amd a butcher" ], "family": "visual_sequence_choice", "index": 7118, "num_...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25955.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25956.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25957.jpg" }, { "path": "../../d...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25955.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25956.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25957.jpg" }, { "path": "../../d...
true
<evidence> E1 | The masked character speaks to a sentry while holding a rifle, and the sentry responds by saying "Oh - Oh!! Maybe a little American double-talk would help..." </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img6.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
multi_image_cot
{ "ground_truth": "E", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "E", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 1291, "num_images": 10, "original_answer": "Answer: Image 5", "p...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img6.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3351-shot1_0/img6.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
true
<evidence> E1 | A hand holding chopsticks is lifting a cooked yellow egg block that is positioned on the right side of the rectangular pan. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3440.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3441.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 19305, "num_images": 4, "original_answer": "True", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "M4...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3440.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3441.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3440.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3441.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first image shows a green, unripe fruit on a plant, while the fourth image shows yellow, ripe fruits on a plant. </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1202.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1203.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1204.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1205.jpg" } ]
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "Dr. Stalvey", "mr. Stalvey", "Miss Stalvey", "Mrs. Stalvey" ], "family": "document_ocr_choice", "index": 4125, "num_images": 4, "original_answer": "mr. Stalvey", "p1_version": "p...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1202.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1203.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1204.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1205.jpg" } ]
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1202.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1203.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1204.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1205.jpg" } ]
true
<evidence> E1 | The "M" field on the "WHILE YOU WERE OUT" slip is handwritten as "Mr. Stalvey". </evidence>
multi_image/cot_objective_multitask_p1_p0_fixed
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img2.jpg" }, { "path": "../../data/images/mantis/im...
multi_image_cot
{ "ground_truth": "G", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "G", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 7056, "num_images": 10, "original_answer": "Answer: Image 7", "p...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img2.jpg" }, { "path": "../../data/images/mantis/im...
[ { "content": "You are a multi-image visual reasoning assistant.\nInspect the provided images and the question carefully. Use only visual facts that are relevant and grounded in the images.\nFirst write a short evidence-based reasoning trace, then give the final answer.\n\nAlways use exactly this output structur...
[ { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_6VPbwxFsO0g-shot36_0/img2.jpg" }, { "path": "../../data/images/mantis/im...
false
End of preview. Expand in Data Studio

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

MIREL training data

The release contains processed parquet files, not the original image archives. Image references in the parquet files are repository-relative and contain no machine-specific usernames or absolute internal paths. Put the downloaded image archives under final/data/images/ (or rewrite the paths as described below) before training.

SFT data

SFT/verified_cot.parquet contains 2,698 unique, verified structured-CoT examples. It is deduplicated by task_uid from the historical 5,396-row file; no example is intentionally repeated in this release. The source tasks are:

  • M4-MIT-States_PropertyCoherence: 1,054 rows;
  • M4-MIT-States_StateCoherence: 594 rows;
  • Mantis-dreamsim: 638 rows;
  • Mantis-imagecode: 412 rows.

The messages include the system prompt used for CoT-SFT and set enable_thinking=false.

PG-OPD data

The two model directories intentionally contain model-specific data; they are not interchangeable.

Qwen3.5-4B

Qwen35-4B/pg_opd.parquet contains 2,394 rows:

Qwen35-4B/pg_opd_8k.parquet contains 8,000 rows for the data-scaling ablation (4,000 verified privilege rows and 4,000 native-control rows). This is an auxiliary scaling artifact.

  • 1,197 verified evidence-privileged rows;
  • 1,197 native fallback rows;
  • evidence was selected from the 4B Stage-A filtering run.

Qwen3.5-9B

Qwen35-9B/pg_opd.parquet contains 3,000 rows:

  • 1,500 verified privilege rows generated and blind-verified with the 9B SFT-LoRA checkpoint;
  • 1,500 native-control rows matched from the same P1 pool;
  • source Stage-A pool: 8,000 P1 samples, with 1,735 verified privilege rows available before the deterministic 1,500-row selection.

For both recipes, privilege text contains evidence content only: no GT answer, answer label, or support=[...] locator is passed to the privileged teacher. Rows retain the native prompt/images and the teacher-side evidence fields. The teacher advantage is computed online by comparing fixed native and privileged teachers on the same student response tokens. There is no GRPO answer reward in PG-OPD.

Original data to download

M4-Instruct

Download the M4-Instruct release described by the official M4-Instruct data card and place the following archives under one local M4-Instruct-Data directory:

  • MIT-States_PropertyCoherence.zip
  • MIT-States_StateCoherence.zip
  • COMICS_Dialogue.zip (needed by the P1 objective-data construction)
  • DocVQA.zip
  • OCR-VQA.zip

The M4 data card and license are included in the upstream M4-Instruct release; the project copy records CC BY 4.0. The processed SFT file uses the two MIT-States subsets. The historical P1 objective construction additionally uses the COMICS, DocVQA and OCR-VQA subsets.

Mantis-Instruct

Download the upstream TIGER-Lab/Mantis-Instruct release and its image archives. The processed files use these Mantis subsets:

  • Mantis-dreamsim for verified SFT;
  • Mantis-imagecode for verified SFT and objective PG/GRPO construction.

Use the upstream dataset's script revision when downloading image archives, as its dataset card explains that the images are provided separately. Follow the upstream license and citation requirements.

Official references:

Image layout and path rewriting

The sanitized parquet paths use this layout relative to the PG-OPD working directory:

../../data/images/m4/...
../../data/images/mantis/...

Place the M4 and Mantis image archives below final/data/images/m4/ and final/data/images/mantis/, preserving the suffixes recorded in the parquet files. Alternatively, use code/PG-OPD/scripts/rewrite_image_paths.py to map ../../data/images to another local image root. The helper changes only image path fields; row order and textual fields remain unchanged. Run its --check-only mode before training.

Downloads last month
13