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/27645.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27646.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27647.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "wise one still breathe", "that will show you who the gods favor my new gun will make me famous among all the apaches !", "evil spirits have struck already at me .", "you will get well , wise...
[ { "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/27645.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27646.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27647.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/27645.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27646.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27647.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character in the yellow tunic is holding a long gun over his shoulder while speaking to the man in blue. </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/3644.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3645.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": 17458, "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/3644.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3645.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/3644.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3645.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | Both images display fence structures with vertical posts and horizontal rails. E2 | Both images show large buildings with prominent dome roofs. </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/OCR-VQA/OCR-VQA/full/images/2833.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2834.jpg" } ]
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "not the same", "Teen & Young Adult", "Children's Books", "Crafts, Hobbies & Home" ], "family": "book_ocr_choice", "index": 5314, "num_images": 2, "original_answer": "Children's Boo...
[ { "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/OCR-VQA/OCR-VQA/full/images/2833.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2834.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/OCR-VQA/OCR-VQA/full/images/2833.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2834.jpg" } ]
true
<evidence> E1 | Both book covers feature prominent titles and author names, indicating they are both published children's books. </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_video4771-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video4771-shot1_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
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": 749, "num_images": 10, "original_answer": "Answer: Image 8", "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_video4771-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video4771-shot1_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_video4771-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video4771-shot1_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
true
<evidence> E1 | The timestamp displayed in the bottom-left corner of the image is 27.360. </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/24420.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24421.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24422.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "as the fighting yank presses self starter", "better get him home so we can patch up that gash on his head !", "if if i ad got into my cart", "but they ' re not going to stop you . as long as...
[ { "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/24420.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24421.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24422.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/24420.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24421.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24422.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character in the green cape is speaking to the man with the bandaged head, suggesting to stay there while he gets the car. </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-videochristmas_3XsTNti1jL8-shot6_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_3XsTNti1jL8-shot6_1/img51.jpg" }, { "path": "../../data/images/manti...
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": 312, "num_images": 10, "original_answer": "Answer: G", "p1_versi...
[ { "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-videochristmas_3XsTNti1jL8-shot6_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_3XsTNti1jL8-shot6_1/img51.jpg" }, { "path": "../../data/images/manti...
[ { "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-videochristmas_3XsTNti1jL8-shot6_1/img48.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_3XsTNti1jL8-shot6_1/img51.jpg" }, { "path": "../../data/images/manti...
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_StateCoherence/MIT-States_StateCoherence/full/images/3812.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3813.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 7422, "num_images": 4, "original_answer": "False", "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/3812.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3813.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/3812.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3813.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
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/3060.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/3061.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-S...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "property_consistency_choice", "index": 16635, "num_images": 4, "original_answer": "False", "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/3060.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/3061.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/3060.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/3061.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/7300.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7301.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7302.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "something is fishy here ! this never happened before", "command you to move !", "can ' t you hear me ? execute the prisoners what ' s the matter ? why don ' t you move ?", "l command you to ...
[ { "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/7300.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7301.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7302.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/7300.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7301.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7302.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/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1100.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1101.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": 8963, "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/1100.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1101.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/1100.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1101.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
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/OCR-VQA/OCR-VQA/full/images/2955.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2956.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "2017", "not the same", "2015", "2016" ], "family": "book_ocr_choice", "index": 7098, "num_images": 2, "original_answer": "2016", "p1_version": "plain_objective_multitask_v1_p0_fi...
[ { "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/OCR-VQA/OCR-VQA/full/images/2955.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2956.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/OCR-VQA/OCR-VQA/full/images/2955.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2956.jpg" } ]
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_StateCoherence/MIT-States_StateCoherence/full/images/1460.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1461.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": 9230, "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/1460.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1461.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/1460.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1461.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | All four images display objects in a static, stationary state. </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/OCR-VQA/OCR-VQA/full/images/3440.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/3441.jpg" } ]
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "not the same", "Humor & Entertainment", "Children's Books", "Computers & Technology" ], "family": "book_ocr_choice", "index": 16327, "num_images": 2, "original_answer": "Children'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/OCR-VQA/OCR-VQA/full/images/3440.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/3441.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/OCR-VQA/OCR-VQA/full/images/3440.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/3441.jpg" } ]
true
<evidence> E1 | Both images depict book covers featuring a central character (an astronaut and a hockey player) surrounded by decorative borders, consistent with the format of children's picture books. </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_fy68dYUZZrc-shot97_1/img33.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot97_1/img35.jpg" }, { "path": "../../data/images/mantis/...
multi_image_cot
{ "ground_truth": "I", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "I", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 18456, "num_images": 10, "original_answer": "Answer: I", "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-videocamping_fy68dYUZZrc-shot97_1/img33.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot97_1/img35.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_fy68dYUZZrc-shot97_1/img33.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot97_1/img35.jpg" }, { "path": "../../data/images/mantis/...
true
<evidence> E1 | The woman is holding and looking at a loaf of bread in her hands. E2 | The child in the green shirt has their hands resting on the table, not touching the plates. </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/2908.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2909.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 7371, "num_images": 4, "original_answer": "False", "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/2908.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2909.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/2908.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2909.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
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/5092.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5093.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5094.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5095.jpg" }, ...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "620", "618", "600", "615" ], "family": "document_ocr_choice", "index": 11993, "num_images": 5, "original_answer": "618", "p1_version": "plain_objective_multitask_v1_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/5092.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5093.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5094.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5095.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/5092.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5093.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5094.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/5095.jpg" }, ...
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/MSR-VTT-videoTrainValVideo_video3339-shot1_4/img218.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3339-shot1_4/img223.jpg" }, { "path": "../../data/images/mantis/imagecode/trai...
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": 10552, "num_images": 10, "original_answer": "Answer: Image 4", "...
[ { "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_video3339-shot1_4/img218.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3339-shot1_4/img223.jpg" }, { "path": "../../data/images/mantis/imagecode/trai...
[ { "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_video3339-shot1_4/img218.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3339-shot1_4/img223.jpg" }, { "path": "../../data/images/mantis/imagecode/trai...
true
<evidence> E1 | A person's left arm extends across the frame, partially obscuring their right hand which is positioned in the center of the image mixing ingredients in a bowl. </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/24505.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24506.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24507.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "follow me ,", "for the next guard and after that the ext ... until i have visited very gate and given them all the pleasure of my company !", "there ' s taylor ! grab him", "here come a coup...
[ { "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/24505.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24506.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24507.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/24505.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24506.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24507.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character in a hat is shown speaking to a guard in a blue uniform. </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-videowedding_4nAse9fEOww-shot41_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_4nAse9fEOww-shot41_0/img5.jpg" }, { "path": "../../data/images/mantis/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": 11123, "num_images": 10, "original_answer": "Answer: Image 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/mantis/imagecode/train/image-sets/video-storytelling-videowedding_4nAse9fEOww-shot41_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_4nAse9fEOww-shot41_0/img5.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-videowedding_4nAse9fEOww-shot41_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_4nAse9fEOww-shot41_0/img5.jpg" }, { "path": "../../data/images/mantis/im...
true
<evidence> E1 | A blonde woman wearing sunglasses is visible on the far left edge of the frame, with her nose cut off by the image boundary. </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/open-images-2309_036a63e2c3ffcd44/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2309_036a63e2c3ffcd44/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
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": 975, "num_images": 10, "original_answer": "Answer: D", "p1_versi...
[ { "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/open-images-2309_036a63e2c3ffcd44/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2309_036a63e2c3ffcd44/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
[ { "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/open-images-2309_036a63e2c3ffcd44/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2309_036a63e2c3ffcd44/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
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/4925.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4926.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4927.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "were both shrinking !", "help we ' re getting smaller and smaller !", "thats all i wanted to know !", "now r can fight back man to man" ], "family": "visual_sequence_choice", "index": ...
[ { "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/4925.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4926.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4927.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/4925.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4926.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4927.jpg" }, { "path": "../../data...
true
<evidence> E1 | The characters are depicted with their heads and bodies visibly reduced in size, accompanied by a speech bubble stating "ME, TOO... LIKE I'M SHRINKING!" </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/7905.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7906.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7907.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "lets not take any chances well dash over the barbed wire fence", "lets go jinx !", "theres the n looks vacant just the same , we ' d better be careful", "well theres a customer for you , jin...
[ { "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/7905.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7906.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7907.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/7905.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7906.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7907.jpg" }, { "path": "../../data...
true
<evidence> E1 | The characters are standing in a barren courtyard with a single building visible in the distance. </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/1085.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1086.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1087.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "going into a huddle with mr . rogers to decide the cast : if you want to wait for the results can we won ' t be long !", "harriet ! harriet ! wait up for us !", "come back honey", "the coach...
[ { "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/1085.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1086.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1087.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/1085.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1086.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1087.jpg" }, { "path": "../../data...
true
<evidence> E1 | A group of people are standing in a circle and talking to a man in a suit. </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/7725.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7726.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7727.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "you clumsy fools you ' ve destroyed my masterpiece !", "keep -- keep him away from me he ' s mad !", "they ' re alive !", "all my work for nothing she asked for the birds she wanted somethin...
[ { "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/7725.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7726.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7727.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/7725.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7726.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/7727.jpg" }, { "path": "../../data...
true
<evidence> E1 | A character reacts with shock and says "CHIRP... CHIRP... CHIRRP!" while a bird is shown flying next to them. </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/1605.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1606.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1607.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "hurry up get him into the car ! we ' ve got to dress him up for the contest", "well hows that for a disguise he ' s a poor kid from the slums who ' s never had a chance ! a musical genius see i ' m"...
[ { "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/1605.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1606.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1607.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/1605.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1606.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/1607.jpg" }, { "path": "../../data...
true
<evidence> E1 | A character in a hat and green jacket is gesturing towards a group of people standing next to a car. </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_fy68dYUZZrc-shot116_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot116_0/img1.jpg" }, { "path": "../../data/images/mantis/...
multi_image_cot
{ "ground_truth": "I", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "I", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 19052, "num_images": 10, "original_answer": "Answer: Image 9", "...
[ { "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_fy68dYUZZrc-shot116_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot116_0/img1.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_fy68dYUZZrc-shot116_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_fy68dYUZZrc-shot116_0/img1.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/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_Bb_nosd2RMo-shot49_1/img62.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_Bb_nosd2RMo-shot49_1/img74.jpg" }, { "path": "../../data/images/man...
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": 10461, "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-videochristmas_Bb_nosd2RMo-shot49_1/img62.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_Bb_nosd2RMo-shot49_1/img74.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-videochristmas_Bb_nosd2RMo-shot49_1/img62.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_Bb_nosd2RMo-shot49_1/img74.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_StateCoherence/MIT-States_StateCoherence/full/images/5748.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/5749.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 16492, "num_images": 4, "original_answer": "False", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "M...
[ { "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/5748.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/5749.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/5748.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/5749.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
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/open-images-2424_05f04546e01499f2/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2424_05f04546e01499f2/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
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": 17672, "num_images": 10, "original_answer": "Answer: H", "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/open-images-2424_05f04546e01499f2/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2424_05f04546e01499f2/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
[ { "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/open-images-2424_05f04546e01499f2/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2424_05f04546e01499f2/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
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/18760.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18761.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18762.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "now to get back to my planet and my carrots --", "carrots here i come", "so far so good !", "next time ling wants to tackle me hed better do it before lunch !" ], "family": "visual_seque...
[ { "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/18760.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18761.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18762.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/18760.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18761.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18762.jpg" }, { "path": "../../d...
true
<evidence> E1 | The rabbit character is smiling and gesturing confidently while saying "B-B-BUT---" and "Your planet is over there! I'm going to throw you in that direction!" </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_video5193-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5193-shot1_0/img4.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": 17481, "num_images": 10, "original_answer": "Answer: E", "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/MSR-VTT-videoTrainValVideo_video5193-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5193-shot1_0/img4.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_video5193-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5193-shot1_0/img4.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
true
<evidence> E1 | The text "METEOROLOGY MC" is partially cut off, showing only the letter "M" on the left side of the frame. </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/1560.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1561.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "state_consistency_choice", "index": 11447, "num_images": 4, "original_answer": "False", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_id": "M...
[ { "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/1560.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1561.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/1560.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/1561.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
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-videochristmas_9pMCYohnRYM-shot158_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_9pMCYohnRYM-shot158_0/img24.jpg" }, { "path": "../../data/images/ma...
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": 18964, "num_images": 10, "original_answer": "Answer: Image 7", "...
[ { "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-videochristmas_9pMCYohnRYM-shot158_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_9pMCYohnRYM-shot158_0/img24.jpg" }, { "path": "../../data/images/ma...
[ { "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-videochristmas_9pMCYohnRYM-shot158_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_9pMCYohnRYM-shot158_0/img24.jpg" }, { "path": "../../data/images/ma...
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/29100.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29101.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29102.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "this bus thank you n for keeping me from being blown across the street", "or i ' d never have caught it !", "this book isn ' t supposed to leave the library lil mouse but ll do you a special fav...
[ { "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/29100.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29101.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29102.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/29100.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29101.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/29102.jpg" }, { "path": "../../d...
true
<evidence> E1 | The characters are standing on a bus stop platform next to a bus with a destination sign, and the dialogue bubble explicitly mentions "L- LET ME GET ON TO THE BUS, LIL FOXES?" and "AS LONG AS YOU DON'T CROSS THE STREET!" </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/10465.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10466.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10467.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "whatever did this can ' t be far away ! i ' ll . signal isobel !", "the werewolf ! lash helpi", "you ' ll never capture me , you vixen !", "where have i heard that voice" ], "family": "v...
[ { "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/10465.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10466.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10467.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/10465.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10466.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/10467.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character in a red suit is running away from a doorway while looking back, and the ground shows a yellow lightning bolt symbol indicating a signal being sent. </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/9155.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9156.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9157.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "holy smoke , easy ! look below !! there ' s a leak !", "well , friend , the next thing on the program will be a forced r landing", "strange the gasoline gauge must be out of order . it shows nea...
[ { "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/9155.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9156.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9157.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/9155.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9156.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/9157.jpg" }, { "path": "../../data...
true
<evidence> E1 | The airplane's fuel gauge on the dashboard is clearly visible and points to the "empty" or low end of the scale. </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/6455.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6456.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6457.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "why the little jerk . nothing but his shirt !", "now a slug through your head to make sure you ' ve", "police car sirens . now well corner the rat", "boys ! yell surround the neighborhood !"...
[ { "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/6455.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6456.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6457.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/6455.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6456.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6457.jpg" }, { "path": "../../data...
true
<evidence> E1 | A character is aiming a pistol directly at a white, shirtless figure. </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/OCR-VQA/OCR-VQA/full/images/1703.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1704.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "Travel", "Children's Books", "not the same", "Medical Books" ], "family": "book_ocr_choice", "index": 15930, "num_images": 2, "original_answer": "Travel", "p1_version": "plain_ob...
[ { "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/OCR-VQA/OCR-VQA/full/images/1703.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1704.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/OCR-VQA/OCR-VQA/full/images/1703.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1704.jpg" } ]
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/open-images-2836_4e89ce8ab3811090/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2836_4e89ce8ab3811090/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
multi_image_cot
{ "ground_truth": "J", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "J", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 895, "num_images": 10, "original_answer": "Answer: Image 10", "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/open-images-2836_4e89ce8ab3811090/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2836_4e89ce8ab3811090/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
[ { "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/open-images-2836_4e89ce8ab3811090/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2836_4e89ce8ab3811090/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-2...
true
<evidence> E1 | The image depicts a black and white line drawing of a cat in the foreground with a small mouse visible in the background. </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/4365.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4366.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4367.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "shouldn ' t flash your gun like that", "aah , ya on ' y nicked me , chum i ' m makin supe my next shot", "you ' re too jitter y yith that roscoe leech .", "that chump was me pal i was just w...
[ { "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/4365.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4366.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4367.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/4365.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4366.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/4367.jpg" }, { "path": "../../data...
true
<evidence> E1 | The speaker questions the reliability of the other parties and expresses a lack of trust in them. </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_video5909-shot0_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5909-shot0_0/img12.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
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": 11858, "num_images": 10, "original_answer": "Answer: F", "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/MSR-VTT-videoTrainValVideo_video5909-shot0_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5909-shot0_0/img12.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
[ { "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_video5909-shot0_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5909-shot0_0/img12.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
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/1676.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1677.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "414 50 8489", "414 50 9849", "414 50 9488", "414 50 9489" ], "family": "document_ocr_choice", "index": 5106, "num_images": 2, "original_answer": "414 50 9489", "p1_version": "pla...
[ { "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/1676.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1677.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/1676.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/1677.jpg" } ]
true
<evidence> E1 | The Social Security Number field contains the value 414 50 9489. </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/21070.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21071.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21072.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "it ' s a mob ! i guess our arrival has a stirred up some excitement", "and from what we ' ve seen thus far id say we have a lot to learn", "follow the exploits of", "exploits of every month ...
[ { "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/21070.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21071.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21072.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/21070.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21071.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/21072.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character in the foreground asks about the noise outside while looking over a railing. </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/2968.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2969.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": 16294, "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/2968.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2969.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/2968.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/2969.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first image shows a clean chest of drawers with smooth surfaces, while the second image shows a sideboard with a distressed, weathered finish. </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/27485.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27486.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27487.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "gulp ! i cant figure out what went wrong", "i can you ' re not as good doing things as telling them", "that would have been a homer n any man ' s league", "ah this fun i" ], "family": "v...
[ { "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/27485.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27486.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27487.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/27485.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27486.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27487.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-videocamping_5lPO6xpAi8Y-shot4_6/img233.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_5lPO6xpAi8Y-shot4_6/img235.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": 7383, "num_images": 10, "original_answer": "Answer: G", "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_5lPO6xpAi8Y-shot4_6/img233.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_5lPO6xpAi8Y-shot4_6/img235.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_5lPO6xpAi8Y-shot4_6/img233.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_5lPO6xpAi8Y-shot4_6/img235.jpg" }, { "path": "../../data/images/mantis/...
true
<evidence> E1 | The circular cutout on the blue chair is fully visible and positioned near the top edge of the frame. </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_cBr1FeeNtFg-shot303_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_cBr1FeeNtFg-shot303_0/img3.jpg" }, { "path": "../../data/images/manti...
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": 5893, "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-videobirthday_cBr1FeeNtFg-shot303_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_cBr1FeeNtFg-shot303_0/img3.jpg" }, { "path": "../../data/images/manti...
[ { "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_cBr1FeeNtFg-shot303_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_cBr1FeeNtFg-shot303_0/img3.jpg" }, { "path": "../../data/images/manti...
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/16495.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16496.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16497.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "what are you waving your magic wand for ?", "mm find water on ory acres farm top the flaming dragon from making hings hot for people and get the un to shine for the children ' s picnic !", "what...
[ { "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/16495.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16496.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16497.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/16495.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16496.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/16497.jpg" }, { "path": "../../d...
true
<evidence> E1 | The wizard is holding a scroll containing a list of tasks, including "find water on ory acres" and "get the un to shine for the children's picnic". </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/19335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19337.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "not going the party were you chum", "everything gone wrong but i ' ve got the rest ae the phony dough , and / zl hide out till things cool down", "here ' s some heavy dough chump !", "one fo...
[ { "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/19335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19337.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/19335.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19336.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19337.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/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3048.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3049.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": 10284, "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/3048.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3049.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/3048.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/3049.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first image shows an unglazed, matte grey Buddha head, while the fourth image shows a glazed, white ceramic pot. </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/27030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27032.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "his gun barrel stunned the second .", "very well js / r .", "yes , sm nazis but to ko him s him", "family life under the b top was a never ending thr to her and she loved ever minute of it" ...
[ { "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/27030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27032.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/27030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/27032.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/111.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/112.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "64248-1", "64248-2", "642481", "64248-3" ], "family": "document_ocr_choice", "index": 16480, "num_images": 2, "original_answer": "64248-1", "p1_version": "plain_objective_multita...
[ { "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/111.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/112.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/111.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/112.jpg" } ]
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/MSR-VTT-videoTrainValVideo_video5745-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5745-shot1_0/img37.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
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": 18583, "num_images": 10, "original_answer": "Answer: F", "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/MSR-VTT-videoTrainValVideo_video5745-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5745-shot1_0/img37.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
[ { "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_video5745-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video5745-shot1_0/img37.jpg" }, { "path": "../../data/images/mantis/imagecode/train/i...
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/4416.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4417.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": 11419, "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/4416.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4417.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/4416.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4417.jpg" } ]
true
<evidence> E1 | The word "tomato" appears in the caption of the first image and the text in the second image. </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/25670.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25671.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25672.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "y that does it we ve got an air view of the harbor !", "ja , but which one ?", "nor port is heavily guarded ! we cant get within two miles or the waterfront !", "only a blur ! hens resisting...
[ { "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/25670.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25671.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25672.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/25670.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25671.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25672.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/COMICS_Dialogue/COMICS_Dialogue/full/images/24800.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24801.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24802.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "feel better now ! able to fight but why do you natives risk your lives to help an american ?", "yes defend our homes ,", "brought him down with a bullet in his w heart ,", "what a sap ive be...
[ { "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/24800.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24801.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24802.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/24800.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24801.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/24802.jpg" }, { "path": "../../d...
true
<evidence> E1 | A soldier is shown lying on the ground in a prone position after being struck by a projectile. </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/4360.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4361.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": 8260, "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/4360.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4361.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/4360.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4361.jpg" } ]
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/6695.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6696.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6697.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "this machine ll fly like an airplane dive like a submarine and smash into obstacles like a tank !", "well let ' s get in !", "snappy outfit you got on", "l at the radio controls , the huge m...
[ { "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/6695.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6696.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6697.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/6695.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6696.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6697.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/20690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20692.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "rats these lower dungeons must be filled with them is that their plant am l to be eaten alive -- by them", "coming from that machine ! it is swinging back and forth , but that gleaming edge i - it c...
[ { "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/20690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20692.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/20690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/20692.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character is depicted as a prisoner with a beard, chained to a wooden plank, and looking distressed in a dark environment. </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/4848.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4849.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4850.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "6,420", "6,4250", "6,245", "6,425." ], "family": "document_ocr_choice", "index": 11474, "num_images": 3, "original_answer": "6,425.", "p1_version": "plain_objective_multitask_v1_...
[ { "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/4848.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4849.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4850.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/4848.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4849.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/4850.jpg" } ]
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/4478.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4479.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": 7866, "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/4478.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4479.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/4478.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4479.jpg" } ]
true
<evidence> E1 | Both images depict a quilted, puffer-style winter coat. </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/8530.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8531.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8532.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "why didn ' t i think . of this long ago hell never think of looking for me in here !", "here he comes now", "oh , my goodness ! that stuff smells awful", "what th ' 2 /" ], "family": "vi...
[ { "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/8530.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8531.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8532.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/8530.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8531.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/8532.jpg" }, { "path": "../../data...
true
<evidence> E1 | The character is hiding inside the scarecrow's body, wearing the suit. </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/18095.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18096.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18097.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "holy cow the fuse", "someone has already torn the fuse in two ... and left a message written in lipstick ! old mill road ! red barn i guess that is where we go", "here it is , mister , but cripe...
[ { "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/18095.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18096.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18097.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/18095.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18096.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/18097.jpg" }, { "path": "../../d...
true
<evidence> E1 | The character is depicted sitting on the ground clutching his head in a manner consistent with having just been struck. </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/5995.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5996.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5997.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "warm , and alive , in this dank tomb , a friend he bent to stroke", "this place is haunted", "marty where are you ?", "shadow share . his voice came ina croak" ], "family": "visual_seque...
[ { "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/5995.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5996.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5997.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/5995.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5996.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/5997.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/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video693-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video693-shot1_0/img2.jpg" }, { "path": "../../data/images/mantis/imagecode/train/imag...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 6459, "num_images": 10, "original_answer": "Answer: Image 3", "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_video693-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video693-shot1_0/img2.jpg" }, { "path": "../../data/images/mantis/imagecode/train/imag...
[ { "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_video693-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video693-shot1_0/img2.jpg" }, { "path": "../../data/images/mantis/imagecode/train/imag...
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/5326.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5327.jpg" } ]
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "painted", "ripe" ], "family": "property_consistency_choice", "index": 3641, "num_images": 2, "original_answer": "ripe", "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/5326.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5327.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/5326.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5327.jpg" } ]
true
<evidence> E1 | the visual content shows a smooth, uniform orange surface, while the visual content shows a textured surface with a green stem and leaf. </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/19030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19032.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "ow , ree", "glad to see you boys helping each other get up in the world !", "one two three .", "okay , beans now . when count three" ], "family": "visual_sequence_choice", "index": 295...
[ { "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/19030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19032.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/19030.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19031.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19032.jpg" }, { "path": "../../d...
true
<evidence> E1 | The boy in the polka-dot pajamas is shouting a command while holding a rope, and the two other characters are pulling on the same rope in response. </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_9KcYGoNCOeE-shot17_1/img125.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_9KcYGoNCOeE-shot17_1/img127.jpg" }, { "path": "../../data/images/manti...
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": 2971, "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_9KcYGoNCOeE-shot17_1/img125.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_9KcYGoNCOeE-shot17_1/img127.jpg" }, { "path": "../../data/images/manti...
[ { "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_9KcYGoNCOeE-shot17_1/img125.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videocamping_9KcYGoNCOeE-shot17_1/img127.jpg" }, { "path": "../../data/images/manti...
true
<evidence> E1 | The woman's face is angled such that the tip of her nose and nostril are visible, while she holds a spoon in front of her arm and the vehicle tire. </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_3l0XYe5kjs4-shot38_3/img229.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_3l0XYe5kjs4-shot38_3/img236.jpg" }, { "path": "../../data/images/man...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 4623, "num_images": 10, "original_answer": "Answer: Image 3", "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_3l0XYe5kjs4-shot38_3/img229.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_3l0XYe5kjs4-shot38_3/img236.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_3l0XYe5kjs4-shot38_3/img229.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videobirthday_3l0XYe5kjs4-shot38_3/img236.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/DocVQA/DocVQA/full/images/891.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/892.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 4, "choice_list": [ "108.94", "109.00", "108.93", "108.95" ], "family": "document_ocr_choice", "index": 15225, "num_images": 2, "original_answer": "108.94", "p1_version": "plain_objective_multitask_v...
[ { "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/891.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/892.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/891.jpg" }, { "path": "../../data/images/m4/DocVQA/DocVQA/full/images/892.jpg" } ]
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/4980.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4981.jpg" } ]
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 2, "choice_list": [ "True", "False" ], "family": "property_consistency_choice", "index": 2743, "num_images": 2, "original_answer": "False", "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/4980.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4981.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/4980.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4981.jpg" } ]
true
<evidence> E1 | the visual content displays a grey shirt with a button-down front, while the visual content displays a blue jacket with an open front. </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/448.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/449.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-Sta...
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": 2890, "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/448.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/449.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-Sta...
[ { "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/448.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/449.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-Sta...
true
<evidence> E1 | All four images display a distinct, decorative frame surrounding a central opening or image. </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/690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/692.jpg" }, { "path": "../../data/im...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "i sure hope t never meet him", "oh , he ' d anyway", "sit tight ... tm going after some water", "im going as far as booming all the time !" ], "family": "visual_sequence_choice", "inde...
[ { "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/690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/692.jpg" }, { "path": "../../data/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/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/690.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/691.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/692.jpg" }, { "path": "../../data/im...
true
<evidence> E1 | The character in the red shirt is standing next to a large truck and speaking to the driver inside the vehicle. </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/4240.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4241.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": 11095, "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/4240.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4241.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/4240.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/4241.jpg" } ]
true
<evidence> E1 | Both images display a vehicle wheel featuring a multi-spoke, metallic silver rim. </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/5880.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5881.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "painted", "ripe" ], "family": "property_consistency_choice", "index": 8219, "num_images": 2, "original_answer": "painted", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_i...
[ { "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/5880.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5881.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/5880.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5881.jpg" } ]
true
<evidence> E1 | The image displays a large, red, three-dimensional letter with a painted surface. </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/OCR-VQA/OCR-VQA/full/images/1012.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1013.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "not the same", "Humor & Entertainment", "Biographies & Memoirs", "Literature & Fiction" ], "family": "book_ocr_choice", "index": 15023, "num_images": 2, "original_answer": "Literat...
[ { "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/OCR-VQA/OCR-VQA/full/images/1012.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1013.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/OCR-VQA/OCR-VQA/full/images/1012.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/1013.jpg" } ]
true
<evidence> E1 | Both images display book covers featuring artistic photography and visual themes consistent with creative works. </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_video3844-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3844-shot1_0/img8.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": 7817, "num_images": 10, "original_answer": "Answer: Image 4", "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_video3844-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3844-shot1_0/img8.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_video3844-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video3844-shot1_0/img8.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
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-videowedding_gfDWSjftuA0-shot83_1/img29.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_gfDWSjftuA0-shot83_1/img34.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": 17569, "num_images": 10, "original_answer": "Answer: H", "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-videowedding_gfDWSjftuA0-shot83_1/img29.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_gfDWSjftuA0-shot83_1/img34.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-videowedding_gfDWSjftuA0-shot83_1/img29.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videowedding_gfDWSjftuA0-shot83_1/img34.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/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot4_2/img169.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot4_2/img175.jpg" }, { "path": "../../data/images/man...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 2209, "num_images": 10, "original_answer": "Answer: Image 2", "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-videochristmas_56Nm66j-i5Q-shot4_2/img169.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot4_2/img175.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-videochristmas_56Nm66j-i5Q-shot4_2/img169.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot4_2/img175.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/OCR-VQA/OCR-VQA/full/images/2185.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2186.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "2015", "not the same", "2017", "2016" ], "family": "book_ocr_choice", "index": 15711, "num_images": 2, "original_answer": "2016", "p1_version": "plain_objective_multitask_v1_p0_f...
[ { "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/OCR-VQA/OCR-VQA/full/images/2185.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2186.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/OCR-VQA/OCR-VQA/full/images/2185.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2186.jpg" } ]
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/25880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25882.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "7 % ae", "but better wood ow the feet than wood the head like soc", "complete data on along the", "england p me dogs /" ], "family": "visual_sequence_choice", "index": 2366, "num_ima...
[ { "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/25880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25882.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/25880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25882.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/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/756.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/757.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohere...
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": 15939, "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/756.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/757.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohere...
[ { "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/756.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/757.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohere...
true
<evidence> E1 | All images depict objects in a static, assembled, and operational state. </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/7368.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7369.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": 10851, "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/7368.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7369.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/7368.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7369.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first image displays a green, unripe fruit on a branch, while the second image displays brown, mature nuts. E2 | The third image displays green, unripe berries on a branch, while the fourth image displays red, ripe berries. </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/6556.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/6557.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": 10420, "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/6556.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/6557.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/6556.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/6557.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first image shows six unpainted metal figures, while the second image shows a single fully painted and detailed figure. </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/OCR-VQA/OCR-VQA/full/images/2060.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2061.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2062.jpg" } ]
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 6, "choice_list": [ "0,1,2", "0,2,1", "1,0,2", "1,2,0", "2,0,1", "2,1,0" ], "family": "book_ocr_choice", "index": 6242, "num_images": 3, "original_answer": "0,2,1", "p1_version": "plain_objec...
[ { "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/OCR-VQA/OCR-VQA/full/images/2060.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2061.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2062.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/OCR-VQA/OCR-VQA/full/images/2060.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2061.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/2062.jpg" } ]
true
<evidence> E1 | The book cover in the visual content is titled "The Truth Seeker" and the book cover in the visual content is titled "The Little Office of the Blessed Virgin Mary". E2 | The book cover in the visual content is titled "Weight Watchers Slow Cooker Recipes". </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/open-images-1488_121dd78597b3dc8d/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1488_121dd78597b3dc8d/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
multi_image_cot
{ "ground_truth": "J", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "J", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 6160, "num_images": 10, "original_answer": "Answer: J", "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/open-images-1488_121dd78597b3dc8d/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1488_121dd78597b3dc8d/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
[ { "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/open-images-1488_121dd78597b3dc8d/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1488_121dd78597b3dc8d/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
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_StateCoherence/MIT-States_StateCoherence/full/images/7132.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7133.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": 12081, "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/7132.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7133.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/7132.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCoherence/full/images/7133.jpg" }, { "path": "../../data/images/m4/MIT-States_StateCoherence/MIT-States_StateCohe...
true
<evidence> E1 | The first set of images shows persimmons transitioning from a green, unripe state to a fully orange, ripe state. E2 | The second set of images shows a pear transitioning from a green, unripe state to a reddish-brown, ripe state. </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/19215.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19216.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19217.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "after the scaly ones are dead shall punish the a mad one , dr , brant !", "come with me there many long noses to", "uta few days later .", "he outwitted e ! but i ' ll ind a way !" ], "f...
[ { "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/19215.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19216.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19217.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/19215.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19216.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/19217.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character in a cape points towards a distant group of figures and says, "It is Jo-Jo! He keeps his word! Look!" </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-videochristmas_56Nm66j-i5Q-shot14_5/img305.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot14_5/img312.jpg" }, { "path": "../../data/images/m...
multi_image_cot
{ "ground_truth": "I", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "I", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 8484, "num_images": 10, "original_answer": "Answer: I", "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-videochristmas_56Nm66j-i5Q-shot14_5/img305.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot14_5/img312.jpg" }, { "path": "../../data/images/m...
[ { "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-videochristmas_56Nm66j-i5Q-shot14_5/img305.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/video-storytelling-videochristmas_56Nm66j-i5Q-shot14_5/img312.jpg" }, { "path": "../../data/images/m...
true
<evidence> E1 | The girl in jeans has a large beige cushion covering her head and upper torso. E2 | The girl in the yellow top is standing on the sofa with her right arm extended, and the hand is blurred and positioned at shoulder height. </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/5034.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5035.jpg" } ]
multi_image_cot
{ "ground_truth": "A", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "A", "choice_count": 2, "choice_list": [ "painted", "ripe" ], "family": "property_consistency_choice", "index": 18140, "num_images": 2, "original_answer": "painted", "p1_version": "plain_objective_multitask_v1_p0_fixed", "row_...
[ { "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/5034.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5035.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/5034.jpg" }, { "path": "../../data/images/m4/MIT-States_PropertyCoherence/MIT-States_PropertyCoherence/full/images/5035.jpg" } ]
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/open-images-12_206a98ccbdce9904/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_206a98ccbdce9904/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_20...
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": 16180, "num_images": 10, "original_answer": "Answer: D", "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/open-images-12_206a98ccbdce9904/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_206a98ccbdce9904/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_20...
[ { "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/open-images-12_206a98ccbdce9904/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_206a98ccbdce9904/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-12_20...
true
<evidence> E1 | The jellyfish are oriented with their bells facing upward and tentacles extending downward against a blue background. E2 | The jellyfish bells display a distinct dark reddish-brown coloration. </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/25385.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25386.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25387.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 4, "choice_list": [ "ath mighty h aww unleashed --", "snappy timing boys !", "he can out . match the strength of elephants !", "let ' s have a little trapeze stuff !" ], "family": "visual_sequence_choice", ...
[ { "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/25385.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25386.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25387.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/25385.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25386.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25387.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character wearing a red shirt and blue pants is speaking to figures in the background while standing near a yellow aircraft. E2 | The character in the red shirt is speaking to two large, elephant-like humanoid figures who are facing away from the viewer. </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/OCR-VQA/OCR-VQA/full/images/4095.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4096.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4097.jpg" } ]
multi_image_cot
{ "ground_truth": "D", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "D", "choice_count": 6, "choice_list": [ "0,1,2", "0,2,1", "1,0,2", "1,2,0", "2,0,1", "2,1,0" ], "family": "book_ocr_choice", "index": 9049, "num_images": 3, "original_answer": "1,2,0", "p1_version": "plain_objec...
[ { "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/OCR-VQA/OCR-VQA/full/images/4095.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4096.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4097.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/OCR-VQA/OCR-VQA/full/images/4095.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4096.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/4097.jpg" } ]
true
<evidence> E1 | The image sequence displays the covers for "The Takings Issue", "Bella the Bunny", and "The Allisons" respectively. </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/6880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6882.jpg" }, { "path": "../../data...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "inside the phone booth .", "everybody here keep calm and nobody will get hurt . don ' t let out a yell , or well have to let you have it", "yeah , baby , how about it shall we ... uh ... why you...
[ { "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/6880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6882.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/6880.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6881.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/6882.jpg" }, { "path": "../../data...
true
<evidence> E1 | The characters are inside a shop, with a woman speaking to men in a tense situation </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/26790.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26791.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26792.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "yow ! somethin ' grabbed me", "its time for my watch at the phones , anyway but tomorrow night that gander sleeps out doors or i do !", "on impulse he gives the nearest toe a hard pinch", "h...
[ { "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/26790.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26791.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26792.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/26790.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26791.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/26792.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/OCR-VQA/OCR-VQA/full/images/872.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/873.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/874.jpg" } ]
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 6, "choice_list": [ "0,1,2", "0,2,1", "1,0,2", "1,2,0", "2,0,1", "2,1,0" ], "family": "book_ocr_choice", "index": 1620, "num_images": 3, "original_answer": "0,2,1", "p1_version": "plain_objec...
[ { "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/OCR-VQA/OCR-VQA/full/images/872.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/873.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/874.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/OCR-VQA/OCR-VQA/full/images/872.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/873.jpg" }, { "path": "../../data/images/m4/OCR-VQA/OCR-VQA/full/images/874.jpg" } ]
true
<evidence> E1 | The cover displays text in the order of "The Definition of Moral Virtue," followed by "The Synchronicity War, Part 4," and finally "Pocket Acapulco: Including Taxco, Ixtapa and Zihuatanejo." </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/25270.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25271.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25272.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "C", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "C", "choice_count": 4, "choice_list": [ "few minutes later as doc ' s vision returns ...", "at the hotel charles", "get off me you mutt !", "a sudden rush !" ], "family": "visual_sequence_choice", "index": 3378, "num_images...
[ { "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/25270.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25271.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25272.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/25270.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25271.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/25272.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/COMICS_Dialogue/COMICS_Dialogue/full/images/23715.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23716.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23717.jpg" }, { "path": "../../d...
multi_image_cot
{ "ground_truth": "B", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "B", "choice_count": 4, "choice_list": [ "a girl - and she ' s caught have to move fast / t wont b hold this robot back forever the main switch but its 700 high to reach", "i hope i can hold / t back long enough for the people to get throug...
[ { "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/23715.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23716.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23717.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/23715.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23716.jpg" }, { "path": "../../data/images/m4/COMICS_Dialogue/COMICS_Dialogue/full/images/23717.jpg" }, { "path": "../../d...
true
<evidence> E1 | A character in a red jacket shouts, "CLEAR THOSE EXITS!!" and another character yells, "HURRY - AS FAST AS YOU CAN!!" </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/open-images-1348_1262edb14d612e0c/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1348_1262edb14d612e0c/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
multi_image_cot
{ "ground_truth": "I", "style": "rule" }
{ "answer_type": "canonical_choice_letter", "canonical_answer": "I", "choice_count": 10, "choice_list": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "family": "image_retrieval", "index": 16667, "num_images": 10, "original_answer": "Answer: I", "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/open-images-1348_1262edb14d612e0c/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1348_1262edb14d612e0c/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
[ { "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/open-images-1348_1262edb14d612e0c/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1348_1262edb14d612e0c/img1.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/open-images-1...
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/MSR-VTT-videoTrainValVideo_video2860-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2860-shot1_0/img5.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": 17097, "num_images": 10, "original_answer": "Answer: Image 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/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2860-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2860-shot1_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_video2860-shot1_0/img0.jpg" }, { "path": "../../data/images/mantis/imagecode/train/image-sets/MSR-VTT-videoTrainValVideo_video2860-shot1_0/img5.jpg" }, { "path": "../../data/images/mantis/imagecode/train/im...
false