Dataset Viewer
Auto-converted to Parquet Duplicate
image
imagewidth (px)
256
868
question
stringclasses
631 values
choices
listlengths
2
7
answer
stringclasses
50 values
answer_type
stringclasses
8 values
auxiliary_images
images listlengths
0
0
capability_name
stringclasses
9 values
question_type_name
stringclasses
7 values
difficulty
stringclasses
3 values
id
stringlengths
17
17
sample_id
stringlengths
51
138
split
stringclasses
1 value
capability_id
stringclasses
9 values
question_type
stringclasses
8 values
template_id
stringclasses
202 values
template_variant_id
stringclasses
999 values
option_labels
listlengths
2
7
option_texts
listlengths
2
7
options_json
stringlengths
19
141
answer_format
stringclasses
2 values
metric_name
stringclasses
3 values
metric_family
stringclasses
3 values
scorer
stringclasses
1 value
scoring_rule
stringclasses
3 values
vlmeval_compatible
bool
1 class
difficulty_reason
stringlengths
49
178
gt_fields
listlengths
1
3
metadata_json
stringlengths
1.6k
5.06k
Which categories belong to the current actionable object set?
[ "A. chefmate 8 frypan", "B. moka pot", "C. white bowl", "D. black book" ]
["A","C"]
multi_choice
visible object list and category recognition
Multiple Choice Question
medium
LIBERO-OBJ-000021
libero_90_KITCHEN_SCENE9_put_the_frying_pan_on_top_of_the_cabinet_seed3_step0003_001659
test
LC1_OBJECT_INVENTORY
multi_choice
LB_C1_MC_OBJECT_SET_10
LB_C1_MC_OBJECT_SET_10_v04
[ "A", "B", "C", "D" ]
[ "chefmate 8 frypan", "moka pot", "white bowl", "black book" ]
{"A":"chefmate 8 frypan","B":"moka pot","C":"white bowl","D":"black book"}
list
Set Exact Match + F1
set_matching
libero_gt_objective_scorer.score_one
Gold and prediction are sets of option letters; report exact match, precision, recall, and F1.
true
Need to perform set membership judgment across multiple candidate categories.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"distractors":["black book","moka pot"],"visible_categories":["white bowl","chefmate 8 frypan"]},"gt_lineage":{"difficulty":"medium","distractor_strategy":"mixed present categories with absent global categories","gt_lineage_depth":2,"involved_gt_edges":["category_membership"],"involved_gt_nodes":["o...
Which category of objects appears in the current scene?
[ "A. black book", "B. glazed rim porcelain ramekin", "C. cream cheese box", "D. bbq sauce" ]
A
single_choice
visible object list and category recognition
Multiple Choice Question
simple
LIBERO-OBJ-000027
libero_90_STUDY_SCENE4_pick_up_the_book_on_the_right_and_place_it_under_the_cabinet_shelf_seed0_step0006_003174
test
LC1_OBJECT_INVENTORY
single_choice
LB_C1_SC_OBJECT_PRESENT_01
LB_C1_SC_OBJECT_PRESENT_01_v01
[ "A", "B", "C", "D" ]
[ "black book", "glazed rim porcelain ramekin", "cream cheese box", "bbq sauce" ]
{"A":"black book","B":"glazed rim porcelain ramekin","C":"cream cheese box","D":"bbq sauce"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
One correct category comes from the current object set, while the rest are categories that do not appear in that scene from the same dataset.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"answer_category":"black book","distractors":["glazed rim porcelain ramekin","cream cheese box","bbq sauce"]},"gt_lineage":{"difficulty":"simple","distractor_strategy":"absent categories from global LIBERO object pool","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":["object:black_bo...
Is the basket in the scene based on the scene label?
[ "A. is", "B. No" ]
B
yes_no
visible object list and category recognition
Binary judgment question
simple
LIBERO-OBJ-000014
libero_90_LIVING_ROOM_SCENE5_put_the_red_mug_on_the_left_plate_seed3_step0006_002430
test
LC1_OBJECT_INVENTORY
yes_no
LB_C1_YN_OBJECT_PRESENT_10
LB_C1_YN_OBJECT_PRESENT_10_v05
[ "A", "B" ]
[ "is", "No" ]
{"A":"is","B":"No"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Existence of a single category is directly inferred from the object's state.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"present":false,"queried_category":"basket"},"gt_lineage":{"difficulty":"simple","distractor_strategy":"same-workspace absent category distractor","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":[]},"involved_objects":[],"provenance":{"episode_idx":3,"gt_fields":["ground_truth.object...
From the object set, how many categories fall into which range?
[ "A. 3-5 classes", "B. 6-8 categories", "C. 9 categories and above", "D. 0-2 classes" ]
D
interval_choice
quantity range and set judgment
Interval Selection Question
medium
LIBERO-OBJ-000000
libero_90_STUDY_SCENE1_pick_up_the_book_and_place_it_in_the_left_compartment_of_the_caddy_seed3_step0006_002718
test
LC2_COUNT_AND_SET
interval_choice
LB_C2_INT_CATEGORY_COUNT_04
LB_C2_INT_CATEGORY_COUNT_04_v05
[ "A", "B", "C", "D" ]
[ "3-5 classes", "6-8 categories", "9 categories and above", "0-2 classes" ]
{"A":"3-5 classes","B":"6-8 categories","C":"9 categories and above","D":"0-2 classes"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The number of category classes after deduplication maps to a fixed range.
[ "ground_truth.objects" ]
{"answer_evidence":{"category_count":2,"interval":"0-2 classes"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"fixed category-count interval bins","gt_lineage_depth":1,"involved_gt_edges":["unique_category_count"],"involved_gt_nodes":["category:black book","category:white yellow mug"]},"involved_objects":[...
This instruction requires handling which objects? Please select all that apply. Task command: pick up the tomato sauce and put it in the tray
[ "A. tomato sauce", "B. butter", "C. alphabet soup" ]
["A"]
multi_choice
quantity range and set judgment
Multiple Choice Question
hard
LIBERO-OBJ-000009
libero_90_LIVING_ROOM_SCENE3_pick_up_the_tomato_sauce_and_put_it_in_the_tray_seed1_step0003_002219
test
LC2_COUNT_AND_SET
multi_choice
LB_C2_MC_TASK_SET_07
LB_C2_MC_TASK_SET_07_v03
[ "A", "B", "C" ]
[ "tomato sauce", "butter", "alphabet soup" ]
{"A":"tomato sauce","B":"butter","C":"alphabet soup"}
list
Set Exact Match + F1
set_matching
libero_gt_objective_scorer.score_one
Gold and prediction are sets of option letters; report exact match, precision, recall, and F1.
true
Bind the natural object names in the task text to the current object category set.
[ "task_language", "ground_truth.objects" ]
{"answer_evidence":{"distractors":["butter","alphabet soup"],"target_categories":["tomato sauce"],"task_language":"pick up the tomato sauce and put it in the tray"},"gt_lineage":{"difficulty":"hard","distractor_strategy":"same-scene non-target objects as distractors","gt_lineage_depth":3,"involved_gt_edges":["task_ment...
In the labeled image, which spatial relation best describes object A relative to object B?
[ "A. right side", "B. front", "C. left side", "D. behind" ]
A
single_choice
Spatial relationships along the three-dimensional axes
Multiple Choice Question
simple
LIBERO-OBJ-000059
libero_90_KITCHEN_SCENE2_put_the_middle_black_bowl_on_top_of_the_cabinet_seed3_step0000_000824
test
LC3_3D_SPATIAL_RELATION
single_choice
LB_C3_SC_AXIS_REL_10
LB_C3_SC_AXIS_REL_10_v01
[ "A", "B", "C", "D" ]
[ "right side", "front", "left side", "behind" ]
{"A":"right side","B":"front","C":"left side","D":"behind"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
choose the signed direction of A relative to B along the x-axis.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"answer_relation":"right side","axis":"x","margin":0.24990555862040692},"gt_lineage":{"answer_margin":0.24990555862040692,"difficulty":"simple","distractor_strategy":"orthogonal-axis relation distractors","gt_lineage_depth":1,"involved_gt_edges":["axis_relation:x"],"involved_gt_nodes":["object:akita...
Please determine the positional relationship between object A and B: Is A lower than B?
[ "A. is", "B. No" ]
B
yes_no
Spatial relationships along the three-dimensional axes
Binary judgment question
medium
LIBERO-OBJ-000004
libero_90_LIVING_ROOM_SCENE1_pick_up_the_cream_cheese_box_and_put_it_in_the_basket_seed1_step0005_001837
test
LC3_3D_SPATIAL_RELATION
yes_no
LB_C3_YN_AXIS_REL_06
LB_C3_YN_AXIS_REL_06_v03
[ "A", "B" ]
[ "is", "No" ]
{"A":"is","B":"No"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the signed difference of the two objects along the z-axis, margin=0.0635.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"A_value":0.5091798360823351,"B_value":0.44569592137963193,"asked_relation":"lower","axis":"z"},"gt_lineage":{"answer_margin":0.06348391470270315,"difficulty":"medium","distractor_strategy":"opposite relation used for false cases","gt_lineage_depth":1,"involved_gt_edges":["axis_relation:z"],"involve...
How far apart are objects A and B? Please choose the interval.
[ "A. greater than or equal to 0.45 meters", "B. 0.08–0.16 meters", "C. 0.28–0.45 meters", "D. 0.16–0.28 meters", "E. less than 0.08 meters" ]
D
interval_choice
distance and proximity relationships
Interval Selection Question
medium
LIBERO-OBJ-000006
libero_90_LIVING_ROOM_SCENE6_put_the_red_mug_on_the_plate_seed1_step0006_002606
test
LC4_DISTANCE_PROXIMITY
interval_choice
LB_C4_INT_PAIR_DISTANCE_08
LB_C4_INT_PAIR_DISTANCE_08_v03
[ "A", "B", "C", "D", "E" ]
[ "greater than or equal to 0.45 meters", "0.08–0.16 meters", "0.28–0.45 meters", "0.16–0.28 meters", "less than 0.08 meters" ]
{"A":"greater than or equal to 0.45 meters","B":"0.08–0.16 meters","C":"0.28–0.45 meters","D":"0.16–0.28 meters","E":"less than 0.08 meters"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
two objects 3D Euclidean distance mapped to a fixed interval.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_m":0.176186,"interval":"0.16–0.28 meters"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"fixed distance interval bins","gt_lineage_depth":2,"involved_gt_edges":["pairwise_distance"],"involved_gt_nodes":["object:porcelain_mug_1","object:red_coffee_mug_1"]},"involved_objects":[{...
With A as a reference, which labeled object is closest to it?
[ "B. label object B", "C. label object C", "D. Label object D" ]
C
marked_object_choice
distance and proximity relationships
question about selecting identified objects in the image
medium
LIBERO-OBJ-000008
libero_goal_put_the_wine_bottle_on_the_rack_seed0_step0003_003491
test
LC4_DISTANCE_PROXIMITY
marked_object_choice
LB_C4_MOC_NEAREST_REF_06
LB_C4_MOC_NEAREST_REF_06_v04
[ "B", "C", "D" ]
[ "label object B", "label object C", "Label object D" ]
{"B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the Euclidean distances of the three candidate objects relative to A.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_to_A":{"B":0.257552,"C":0.125404,"D":0.256284}},"gt_lineage":{"answer_margin":0.13087960717574754,"difficulty":"medium","distractor_strategy":"nearest-neighbor distractors from same scene objects","gt_lineage_depth":2,"involved_gt_edges":["distance(A,candidate)"],"involved_gt_nodes":["obje...
Observe the lateral indicator diagram, which labeled object is taller?
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
C
marked_object_choice
Sorting and Area Localization
question about selecting identified objects in the image
medium
LIBERO-OBJ-000231
libero_90_LIVING_ROOM_SCENE2_pick_up_the_orange_juice_and_put_it_in_the_basket_seed0_step0004_002020
test
LC5_ORDER_AND_REGION
marked_object_choice
LB_C5_MOC_HIGHEST_05
LB_C5_MOC_HIGHEST_05_v02
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare candidate object height z, highest vs second highest margin=0.0340.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"z_by_label":{"A":0.47517012396013675,"B":0.43219202665002604,"C":0.5091796473738082,"D":0.47516915037869295}},"gt_lineage":{"answer_margin":0.034009523413671416,"difficulty":"medium","distractor_strategy":"height comparison with near-tie rejection","gt_lineage_depth":2,"involved_gt_edges":["max_hei...
Please determine the right-to-left sorting order of A, B, C, D from the annotated image.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
["A","D","B","C"]
ordered_list
Sorting and Area Localization
Sorting Question
hard
LIBERO-OBJ-000005
libero_90_KITCHEN_SCENE2_put_the_black_bowl_at_the_front_on_the_plate_seed1_step0001_000745
test
LC5_ORDER_AND_REGION
ordered_list
LB_C5_ORD_AXIS_10
LB_C5_ORD_AXIS_10_v05
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
list
Exact Order Accuracy + Pairwise Accuracy
ranking
libero_gt_objective_scorer.score_one
Exact sequence equality plus Kendall-style pairwise consistency over option letters.
true
Four objects sorted along the x-axis with the minimum adjacent spacing = 0.0674.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"axis":"x","values_by_label":{"A":0.11757005091363121,"B":-0.05330572719407414,"C":-0.14621754963284003,"D":0.014110191494072898}},"gt_lineage":{"answer_margin":0.06741591868814704,"difficulty":"hard","distractor_strategy":"near-tie rejected axis ordering","gt_lineage_depth":2,"involved_gt_edges":["...
Choose the corresponding interval based on the gripper state in the labeled image.
[ "A. closed or nearly closed", "B. open", "C. half-open" ]
C
interval_choice
robot end effector state and action direction
Interval Selection Question
simple
LIBERO-OBJ-000073
libero_90_LIVING_ROOM_SCENE2_pick_up_the_tomato_sauce_and_put_it_in_the_basket_seed2_step0007_002071
test
LC6_ROBOT_EEF_ACTION
interval_choice
LB_C6_INT_GRIPPER_07
LB_C6_INT_GRIPPER_07_v05
[ "A", "B", "C" ]
[ "closed or nearly closed", "open", "half-open" ]
{"A":"closed or nearly closed","B":"open","C":"half-open"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
gripper qpos absolute mean mapped to discrete open-closed range.
[ "ground_truth.robot_state.robot0_gripper_qpos" ]
{"answer_evidence":{"gripper_opening_proxy":0.038321,"interval":"half-open"},"gt_lineage":{"difficulty":"simple","distractor_strategy":"fixed non-numeric gripper state intervals","gt_lineage_depth":1,"involved_gt_edges":["gripper_opening_interval"],"involved_gt_nodes":["robot:gripper_qpos"]},"involved_objects":[],"prov...
Select the candidate object that is closest to the end effector of the robotic arm.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
B
marked_object_choice
robot end effector state and action direction
question about selecting identified objects in the image
medium
LIBERO-OBJ-000058
libero_spatial_pick_up_the_black_bowl_in_the_top_drawer_of_the_wooden_cabinet_and_place_it_on_the_plate_seed3_step0007_003999
test
LC6_ROBOT_EEF_ACTION
marked_object_choice
LB_C6_MOC_EEF_NEAREST_04
LB_C6_MOC_EEF_NEAREST_04_v03
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the distance from EEF to four candidate objects, with the closest distance being 0.0486.
[ "ground_truth.robot_state.robot0_eef_pos", "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_to_eef":{"A":0.356413,"B":0.307789,"C":0.363035,"D":0.417621},"eef_position":[-0.1642039207669719,-0.00045259255994419163,1.1811673489903602]},"gt_lineage":{"answer_margin":0.048623526044553556,"difficulty":"medium","distractor_strategy":"same-scene object candidates; nearest-tie rejected"...
What is the main movement direction for this step action?
[ "A. down", "B. upward", "C. forward", "D. backward", "E. left", "F. right", "G. Very little movement" ]
F
single_choice
robot end effector state and action direction
Multiple Choice Question
medium
LIBERO-OBJ-000007
libero_90_STUDY_SCENE3_pick_up_the_red_mug_and_place_it_to_the_right_of_the_caddy_seed2_step0002_003026
test
LC6_ROBOT_EEF_ACTION
single_choice
LB_C6_SC_ACTION_DIR_01
LB_C6_SC_ACTION_DIR_01_v03
[ "A", "B", "C", "D", "E", "F", "G" ]
[ "down", "upward", "forward", "backward", "left", "right", "Very little movement" ]
{"A":"down","B":"upward","C":"forward","D":"backward","E":"left","F":"right","G":"Very little movement"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The action vector main axis is x, with a margin of 0.1259.
[ "ground_truth.action" ]
{"answer_evidence":{"action":[0.17142857611179352,0.04553571343421936,-0.0,0.0,-0.009642857126891613,0.010714286006987095,-1.0],"axis":"x","direction":"right"},"gt_lineage":{"answer_margin":0.12589286267757416,"difficulty":"medium","distractor_strategy":"dominant-axis action choices; tied vectors rejected","gt_lineage_...
Which annotated object has the largest position change between the first and last frames?
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
A
marked_object_choice
sequence trajectory and state change
question about selecting identified objects in the image
hard
LIBERO-OBJ-000052
libero_spatial_pick_up_the_black_bowl_on_the_ramekin_and_place_it_on_the_plate_seed1_step0006_004014
test
LC7_TEMPORAL_TRAJECTORY
marked_object_choice
LB_C7_MOC_MOVED_OBJECT_05
LB_C7_MOC_MOVED_OBJECT_05_v05
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the displacements of the four objects between the first and last frames of the same episode. There is a clear margin between the largest and second largest displacements.
[ "step", "ground_truth.objects.*.position" ]
{"answer_evidence":{"displacement_by_label":{"A":0.010823,"B":5.6e-05,"C":6.2e-05,"D":6.4e-05},"sequence_key":["libero_spatial","pick_up_the_black_bowl_on_the_ramekin_and_place_it_on_the_plate",1,1]},"gt_lineage":{"answer_margin":0.010757061967333543,"difficulty":"hard","distractor_strategy":"moving-object distractors ...
Observe the final state and determine the order of distances from A, B, C, D to E.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
["A","D","B","C"]
ordered_list
sequence trajectory and state change
Sorting Question
hard
LIBERO-OBJ-000003
libero_10_LIVING_ROOM_SCENE2_put_both_the_cream_cheese_box_and_the_butter_in_the_basket_seed1_step0006_000046
test
LC7_TEMPORAL_TRAJECTORY
ordered_list
LB_C7_ORD_END_DISTANCE_08
LB_C7_ORD_END_DISTANCE_08_v02
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
list
Exact Order Accuracy + Pairwise Accuracy
ranking
libero_gt_objective_scorer.score_one
Exact sequence equality plus Kendall-style pairwise consistency over option letters.
true
First locate the frame where the sequence ends, then sort by the distance from the object to the EEF.
[ "step", "ground_truth.robot_state.robot0_eef_pos", "ground_truth.objects.*.position" ]
{"answer_evidence":{"final_distance_to_eef":{"A":0.204441,"B":0.324768,"C":0.401657,"D":0.273354},"sequence_key":["libero_10","LIVING_ROOM_SCENE2_put_both_the_cream_cheese_box_and_the_butter_in_the_basket",1,1]},"gt_lineage":{"answer_margin":0.05141360711367027,"difficulty":"hard","distractor_strategy":"final-frame nea...
Where did the end of the robotic arm mainly move between the initial and final states?
[ "A. forward", "B. left", "C. upward", "D. backward", "E. Very little movement", "F. down", "G. right" ]
G
temporal_choice
sequence trajectory and state change
Temporal Choice Question
medium
LIBERO-OBJ-000015
libero_90_KITCHEN_SCENE2_open_the_top_drawer_of_the_cabinet_seed3_step0004_000700
test
LC7_TEMPORAL_TRAJECTORY
temporal_choice
LB_C7_TC_EEF_DIR_02
LB_C7_TC_EEF_DIR_02_v05
[ "A", "B", "C", "D", "E", "F", "G" ]
[ "forward", "left", "upward", "backward", "Very little movement", "down", "right" ]
{"A":"forward","B":"left","C":"upward","D":"backward","E":"Very little movement","F":"down","G":"right"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the EEF positions at the beginning and end of the sequence, with the main axis = x and margin = 0.0099.
[ "step", "ground_truth.robot_state.robot0_eef_pos" ]
{"answer_evidence":{"delta":[0.04147765749213278,-0.03158705830802648,0.006672790533202688],"direction":"right","end_eef":[-0.1548108853033539,-0.04305255088409927,1.1698823942220855],"sequence_key":["libero_90","KITCHEN_SCENE2_open_the_top_drawer_of_the_cabinet",3,3],"start_eef":[-0.1962885427954867,-0.011465492576072...
Please select the target object mentioned in the task: pick up the tomato sauce and put it in the basket
[ "A. basket", "B. alphabet soup", "C. cream cheese box", "D. tomato sauce" ]
["A","D"]
multi_choice
task semantics bound to goals
Multiple Choice Question
hard
LIBERO-OBJ-000012
libero_90_LIVING_ROOM_SCENE1_pick_up_the_tomato_sauce_and_put_it_in_the_basket_seed3_step0006_001918
test
LC8_TASK_SEMANTIC_BINDING
multi_choice
LB_C8_MC_TASK_TARGETS_09
LB_C8_MC_TASK_TARGETS_09_v02
[ "A", "B", "C", "D" ]
[ "basket", "alphabet soup", "cream cheese box", "tomato sauce" ]
{"A":"basket","B":"alphabet soup","C":"cream cheese box","D":"tomato sauce"}
list
Set Exact Match + F1
set_matching
libero_gt_objective_scorer.score_one
Gold and prediction are sets of option letters; report exact match, precision, recall, and F1.
true
Bind the natural object names in the task text to the current object category set.
[ "task_language", "ground_truth.objects" ]
{"answer_evidence":{"distractors":["alphabet soup","cream cheese box"],"target_categories":["basket","tomato sauce"],"task_language":"pick up the tomato sauce and put it in the basket"},"gt_lineage":{"difficulty":"hard","distractor_strategy":"same-scene non-target objects as distractors","gt_lineage_depth":3,"involved_...
According to the task description, the robot mainly needs to complete which action? Task instruction: pick up the butter and put it in the basket
[ "A. other operations", "B. close object", "C. pick up object", "D. place/move object", "E. open object", "F. turn on the switch or device" ]
C
single_choice
task semantics bound to goals
Multiple Choice Question
medium
LIBERO-OBJ-000244
libero_90_LIVING_ROOM_SCENE2_pick_up_the_butter_and_put_it_in_the_basket_seed1_step0000_001960
test
LC8_TASK_SEMANTIC_BINDING
single_choice
LB_C8_SC_TASK_ACTION_04
LB_C8_SC_TASK_ACTION_04_v02
[ "A", "B", "C", "D", "E", "F" ]
[ "other operations", "close object", "pick up object", "place/move object", "open object", "turn on the switch or device" ]
{"A":"other operations","B":"close object","C":"pick up object","D":"place/move object","E":"open object","F":"turn on the switch or device"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The task language maps to action families through fixed verb rules.
[ "task_language" ]
{"answer_evidence":{"action_family":"pick up object","task_language":"pick up the butter and put it in the basket"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"other action families as semantic distractors","gt_lineage_depth":1,"involved_gt_edges":["action_family_parse"],"involved_gt_nodes":["task_langua...
Which of the two perspectives, A or B, is the robot arm hand-eye perspective?
[ "A. Panel A", "B. Panel B" ]
A
multi_view_choice
Dual-view Consistency and Camera Recognition
Multi-view Consistency Multiple Choice
simple
LIBERO-OBJ-000001
libero_90_LIVING_ROOM_SCENE5_put_the_yellow_and_white_mug_on_the_right_plate_seed0_step0007_002503
test
LC9_MULTI_VIEW_CONSISTENCY
multi_view_choice
LB_C9_MV_CAMERA_06
LB_C9_MV_CAMERA_06_v02
[ "A", "B" ]
[ "Panel A", "Panel B" ]
{"A":"Panel A","B":"Panel B"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Two synchronized RGB camera sources are randomly arranged, and the answer is uniquely determined by the camera key.
[ "images.agentview_image", "images.robot0_eye_in_hand_image" ]
{"answer_evidence":{"panel_sources":{"A":"robot0_eye_in_hand_image","B":"agentview_image"}},"gt_lineage":{"difficulty":"simple","distractor_strategy":"camera order randomized","gt_lineage_depth":1,"involved_gt_edges":["camera_source_mapping"],"involved_gt_nodes":["image:robot0_eye_in_hand_image","image:agentview_image"...
Which candidate object category belongs to the current scene?
[ "A. black bowl", "B. wooden tray", "C. ketchup", "D. bbq sauce" ]
A
single_choice
visible object list and category recognition
Multiple Choice Question
simple
LIBERO-OBJ-000031
libero_90_KITCHEN_SCENE4_close_the_bottom_drawer_of_the_cabinet_and_open_the_top_drawer_seed0_step0003_001059
test
LC1_OBJECT_INVENTORY
single_choice
LB_C1_SC_OBJECT_PRESENT_05
LB_C1_SC_OBJECT_PRESENT_05_v03
[ "A", "B", "C", "D" ]
[ "black bowl", "wooden tray", "ketchup", "bbq sauce" ]
{"A":"black bowl","B":"wooden tray","C":"ketchup","D":"bbq sauce"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
One correct category comes from the current object set, while the rest are categories that do not appear in that scene from the same dataset.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"answer_category":"black bowl","distractors":["wooden tray","ketchup","bbq sauce"]},"gt_lineage":{"difficulty":"simple","distractor_strategy":"absent categories from global LIBERO object pool","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":["object:akita_black_bowl_1"]},"involved_ob...
Based on the scene identifier, is the moka pot in the scene?
[ "A. is", "B. No" ]
A
yes_no
visible object list and category recognition
Binary judgment question
simple
LIBERO-OBJ-000221
libero_10_KITCHEN_SCENE8_put_both_moka_pots_on_the_stove_seed0_step0005_000261
test
LC1_OBJECT_INVENTORY
yes_no
LB_C1_YN_OBJECT_PRESENT_05
LB_C1_YN_OBJECT_PRESENT_05_v05
[ "A", "B" ]
[ "is", "No" ]
{"A":"is","B":"No"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Existence of a single category is directly inferred from the object's state.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"present":true,"queried_category":"moka pot"},"gt_lineage":{"difficulty":"simple","distractor_strategy":"same-workspace absent category distractor","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":["object:moka_pot_1","object:moka_pot_2"]},"involved_objects":[{"category":"moka pot","d...
How many object categories in the current scene fall into which range?
[ "A. 9 categories and above", "B. 0-2 classes", "C. 3-5 classes", "D. 6-8 categories" ]
B
interval_choice
quantity range and set judgment
Interval Selection Question
medium
LIBERO-OBJ-000002
libero_10_STUDY_SCENE1_pick_up_the_book_and_place_it_in_the_back_compartment_of_the_caddy_seed3_step0005_000189
test
LC2_COUNT_AND_SET
interval_choice
LB_C2_INT_CATEGORY_COUNT_10
LB_C2_INT_CATEGORY_COUNT_10_v01
[ "A", "B", "C", "D" ]
[ "9 categories and above", "0-2 classes", "3-5 classes", "6-8 categories" ]
{"A":"9 categories and above","B":"0-2 classes","C":"3-5 classes","D":"6-8 categories"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The number of category classes after deduplication maps to a fixed range.
[ "ground_truth.objects" ]
{"answer_evidence":{"category_count":2,"interval":"0-2 classes"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"fixed category-count interval bins","gt_lineage_depth":1,"involved_gt_edges":["unique_category_count"],"involved_gt_nodes":["category:black book","category:white yellow mug"]},"involved_objects":[...
This instruction requires handling which objects? Please select all that apply. task instruction: pick up the alphabet soup and put it in the basket
[ "A. tomato sauce", "B. basket", "C. cream cheese box", "D. alphabet soup" ]
["B","D"]
multi_choice
quantity range and set judgment
Multiple Choice Question
hard
LIBERO-OBJ-000118
libero_90_LIVING_ROOM_SCENE1_pick_up_the_alphabet_soup_and_put_it_in_the_basket_seed1_step0001_001801
test
LC2_COUNT_AND_SET
multi_choice
LB_C2_MC_TASK_SET_06
LB_C2_MC_TASK_SET_06_v03
[ "A", "B", "C", "D" ]
[ "tomato sauce", "basket", "cream cheese box", "alphabet soup" ]
{"A":"tomato sauce","B":"basket","C":"cream cheese box","D":"alphabet soup"}
list
Set Exact Match + F1
set_matching
libero_gt_objective_scorer.score_one
Gold and prediction are sets of option letters; report exact match, precision, recall, and F1.
true
Bind the natural object names in the task text to the current object category set.
[ "task_language", "ground_truth.objects" ]
{"answer_evidence":{"distractors":["cream cheese box","tomato sauce"],"target_categories":["alphabet soup","basket"],"task_language":"pick up the alphabet soup and put it in the basket"},"gt_lineage":{"difficulty":"hard","distractor_strategy":"same-scene non-target objects as distractors","gt_lineage_depth":3,"involved...
In the labeled image, which spatial relation best describes object A relative to object B?
[ "A. behind", "B. right side", "C. left side", "D. front" ]
C
single_choice
Spatial relationships along the three-dimensional axes
Multiple Choice Question
simple
LIBERO-OBJ-000116
libero_90_STUDY_SCENE4_pick_up_the_book_on_the_right_and_place_it_on_the_cabinet_shelf_seed3_step0006_003166
test
LC3_3D_SPATIAL_RELATION
single_choice
LB_C3_SC_AXIS_REL_01
LB_C3_SC_AXIS_REL_01_v01
[ "A", "B", "C", "D" ]
[ "behind", "right side", "left side", "front" ]
{"A":"behind","B":"right side","C":"left side","D":"front"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
choose the signed direction of A relative to B along the x-axis.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"answer_relation":"left side","axis":"x","margin":0.05976718698668895},"gt_lineage":{"answer_margin":0.05976718698668895,"difficulty":"simple","distractor_strategy":"orthogonal-axis relation distractors","gt_lineage_depth":1,"involved_gt_edges":["axis_relation:x"],"involved_gt_nodes":["object:yellow...
Please determine the positional relationship between object A and B: is A to the right of B?
[ "A. is", "B. No" ]
B
yes_no
Spatial relationships along the three-dimensional axes
Binary judgment question
medium
LIBERO-OBJ-000011
libero_goal_put_the_wine_bottle_on_the_rack_seed2_step0004_003508
test
LC3_3D_SPATIAL_RELATION
yes_no
LB_C3_YN_AXIS_REL_10
LB_C3_YN_AXIS_REL_10_v03
[ "A", "B" ]
[ "is", "No" ]
{"A":"is","B":"No"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the signed difference of the two objects along the x-axis, margin=0.1573.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"A_value":-0.21117884057685843,"B_value":-0.053922771930130176,"asked_relation":"right side","axis":"x"},"gt_lineage":{"answer_margin":0.15725606864672825,"difficulty":"medium","distractor_strategy":"opposite relation used for false cases","gt_lineage_depth":1,"involved_gt_edges":["axis_relation:x"]...
Which interval does the distance between labeled object A and labeled object B belong to?
[ "A. 0.16–0.28 meters", "B. less than 0.08 meters", "C. greater than or equal to 0.45 meters", "D. 0.28–0.45 meters", "E. 0.08–0.16 meters" ]
A
interval_choice
distance and proximity relationships
Interval Selection Question
medium
LIBERO-OBJ-000017
libero_object_pick_up_the_bbq_sauce_and_place_it_in_the_basket_seed0_step0000_003616
test
LC4_DISTANCE_PROXIMITY
interval_choice
LB_C4_INT_PAIR_DISTANCE_10
LB_C4_INT_PAIR_DISTANCE_10_v01
[ "A", "B", "C", "D", "E" ]
[ "0.16–0.28 meters", "less than 0.08 meters", "greater than or equal to 0.45 meters", "0.28–0.45 meters", "0.08–0.16 meters" ]
{"A":"0.16–0.28 meters","B":"less than 0.08 meters","C":"greater than or equal to 0.45 meters","D":"0.28–0.45 meters","E":"0.08–0.16 meters"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
two objects 3D Euclidean distance mapped to a fixed interval.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_m":0.175295,"interval":"0.16–0.28 meters"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"fixed distance interval bins","gt_lineage_depth":2,"involved_gt_edges":["pairwise_distance"],"involved_gt_nodes":["object:cream_cheese_1","object:chocolate_pudding_1"]},"involved_objects":...
Among the candidate objects in the image, which one is closest to A?
[ "B. label object B", "C. label object C", "D. Label object D" ]
B
marked_object_choice
distance and proximity relationships
question about selecting identified objects in the image
medium
LIBERO-OBJ-000023
libero_90_LIVING_ROOM_SCENE4_stack_the_left_bowl_on_the_right_bowl_and_place_them_in_the_tray_seed2_step0002_002354
test
LC4_DISTANCE_PROXIMITY
marked_object_choice
LB_C4_MOC_NEAREST_REF_03
LB_C4_MOC_NEAREST_REF_03_v05
[ "B", "C", "D" ]
[ "label object B", "label object C", "Label object D" ]
{"B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the Euclidean distances of the three candidate objects relative to A.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_to_A":{"B":0.239482,"C":0.443443,"D":0.393609}},"gt_lineage":{"answer_margin":0.1541269389282019,"difficulty":"medium","distractor_strategy":"nearest-neighbor distractors from same scene objects","gt_lineage_depth":2,"involved_gt_edges":["distance(A,candidate)"],"involved_gt_nodes":["objec...
Observe the diagram and arrange A, B, C, D from front to back.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
["B","D","C","A"]
ordered_list
Sorting and Area Localization
Sorting Question
hard
LIBERO-OBJ-000018
libero_90_LIVING_ROOM_SCENE5_put_the_white_mug_on_the_left_plate_seed3_step0002_002490
test
LC5_ORDER_AND_REGION
ordered_list
LB_C5_ORD_AXIS_05
LB_C5_ORD_AXIS_05_v02
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
list
Exact Order Accuracy + Pairwise Accuracy
ranking
libero_gt_objective_scorer.score_one
Exact sequence equality plus Kendall-style pairwise consistency over option letters.
true
Four objects sorted by y-axis, minimum adjacent spacing = 0.0844.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"axis":"y","values_by_label":{"A":-0.2876134987558727,"B":0.09987282526325358,"C":-0.15609503912447895,"D":0.015513664159950001}},"gt_lineage":{"answer_margin":0.08435916110330358,"difficulty":"hard","distractor_strategy":"near-tie rejected axis ordering","gt_lineage_depth":2,"involved_gt_edges":["s...
Choose the corresponding interval based on the gripper state in the labeled image.
[ "A. open", "B. closed or nearly closed", "C. half-open" ]
C
interval_choice
robot end effector state and action direction
Interval Selection Question
simple
LIBERO-OBJ-000165
libero_90_KITCHEN_SCENE5_put_the_black_bowl_on_the_plate_seed3_step0000_001304
test
LC6_ROBOT_EEF_ACTION
interval_choice
LB_C6_INT_GRIPPER_02
LB_C6_INT_GRIPPER_02_v05
[ "A", "B", "C" ]
[ "open", "closed or nearly closed", "half-open" ]
{"A":"open","B":"closed or nearly closed","C":"half-open"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
gripper qpos absolute mean mapped to discrete open-closed range.
[ "ground_truth.robot_state.robot0_gripper_qpos" ]
{"answer_evidence":{"gripper_opening_proxy":0.030968,"interval":"half-open"},"gt_lineage":{"difficulty":"simple","distractor_strategy":"fixed non-numeric gripper state intervals","gt_lineage_depth":1,"involved_gt_edges":["gripper_opening_interval"],"involved_gt_nodes":["robot:gripper_qpos"]},"involved_objects":[],"prov...
Select the candidate object that is closest to the end effector of the robotic arm.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
B
marked_object_choice
robot end effector state and action direction
question about selecting identified objects in the image
medium
LIBERO-OBJ-000109
libero_90_KITCHEN_SCENE2_put_the_black_bowl_at_the_back_on_the_plate_seed0_step0006_000710
test
LC6_ROBOT_EEF_ACTION
marked_object_choice
LB_C6_MOC_EEF_NEAREST_01
LB_C6_MOC_EEF_NEAREST_01_v03
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the distance from EEF to four candidate objects, with the closest distance being 0.0746.
[ "ground_truth.robot_state.robot0_eef_pos", "ground_truth.objects.*.position" ]
{"answer_evidence":{"distance_to_eef":{"A":0.340993,"B":0.266437,"C":0.359311,"D":0.438519},"eef_position":[-0.2259460892230951,0.007868799253929564,1.1556884703811483]},"gt_lineage":{"answer_margin":0.07455518695810898,"difficulty":"medium","distractor_strategy":"same-scene object candidates; nearest-tie rejected","gt...
Which annotated object has the largest position change between the first and last frames?
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
D
marked_object_choice
sequence trajectory and state change
question about selecting identified objects in the image
hard
LIBERO-OBJ-000060
libero_10_LIVING_ROOM_SCENE1_put_both_the_alphabet_soup_and_the_cream_cheese_box_in_the_basket_seed3_step0002_000250
test
LC7_TEMPORAL_TRAJECTORY
marked_object_choice
LB_C7_MOC_MOVED_OBJECT_02
LB_C7_MOC_MOVED_OBJECT_02_v05
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the displacements of the four objects between the first and last frames of the same episode. There is a clear margin between the largest and second largest displacements.
[ "step", "ground_truth.objects.*.position" ]
{"answer_evidence":{"displacement_by_label":{"A":0.0,"B":0.0,"C":0.0,"D":0.083111},"sequence_key":["libero_10","LIVING_ROOM_SCENE1_put_both_the_alphabet_soup_and_the_cream_cheese_box_in_the_basket",3,3]},"gt_lineage":{"answer_margin":0.08311036501060433,"difficulty":"hard","distractor_strategy":"moving-object distracto...
Observe the final state and determine the order of distances from A, B, C, D to E.
[ "A. label object A", "B. label object B", "C. label object C", "D. Label object D" ]
["A","B","C","D"]
ordered_list
sequence trajectory and state change
Sorting Question
hard
LIBERO-OBJ-000019
libero_90_LIVING_ROOM_SCENE3_pick_up_the_alphabet_soup_and_put_it_in_the_tray_seed2_step0006_002102
test
LC7_TEMPORAL_TRAJECTORY
ordered_list
LB_C7_ORD_END_DISTANCE_10
LB_C7_ORD_END_DISTANCE_10_v02
[ "A", "B", "C", "D" ]
[ "label object A", "label object B", "label object C", "Label object D" ]
{"A":"label object A","B":"label object B","C":"label object C","D":"Label object D"}
list
Exact Order Accuracy + Pairwise Accuracy
ranking
libero_gt_objective_scorer.score_one
Exact sequence equality plus Kendall-style pairwise consistency over option letters.
true
First locate the frame where the sequence ends, then sort by the distance from the object to the EEF.
[ "step", "ground_truth.robot_state.robot0_eef_pos", "ground_truth.objects.*.position" ]
{"answer_evidence":{"final_distance_to_eef":{"A":0.217723,"B":0.27253,"C":0.303088,"D":0.353336},"sequence_key":["libero_90","LIVING_ROOM_SCENE3_pick_up_the_alphabet_soup_and_put_it_in_the_tray",2,2]},"gt_lineage":{"answer_margin":0.030557352180704878,"difficulty":"hard","distractor_strategy":"final-frame near-to-far o...
Where did the end of the robotic arm mainly move between the initial and final states?
[ "A. down", "B. backward", "C. Very little movement", "D. forward", "E. right", "F. left", "G. upward" ]
B
temporal_choice
sequence trajectory and state change
Temporal Choice Question
medium
LIBERO-OBJ-000026
libero_90_LIVING_ROOM_SCENE2_pick_up_the_orange_juice_and_put_it_in_the_basket_seed2_step0002_002034
test
LC7_TEMPORAL_TRAJECTORY
temporal_choice
LB_C7_TC_EEF_DIR_07
LB_C7_TC_EEF_DIR_07_v05
[ "A", "B", "C", "D", "E", "F", "G" ]
[ "down", "backward", "Very little movement", "forward", "right", "left", "upward" ]
{"A":"down","B":"backward","C":"Very little movement","D":"forward","E":"right","F":"left","G":"upward"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Compare the EEF positions at the beginning and end of the sequence, main axis=y, margin=0.0374.
[ "step", "ground_truth.robot_state.robot0_eef_pos" ]
{"answer_evidence":{"delta":[0.014882982468190087,-0.05223636726529239,-0.0033677111400828297],"direction":"backward","end_eef":[-0.028198121370677745,-0.07038443658071204,0.6861026723972121],"sequence_key":["libero_90","LIVING_ROOM_SCENE2_pick_up_the_orange_juice_and_put_it_in_the_basket",2,2],"start_eef":[-0.04308110...
According to the task instruction, which candidate objects need to be manipulated? Select all that apply. Task instruction: pick up the bbq sauce and place it in the basket
[ "A. basket", "B. ketchup", "C. bbq sauce", "D. salad dressing" ]
["A","C"]
multi_choice
task semantics bound to goals
Multiple Choice Question
hard
LIBERO-OBJ-000016
libero_object_pick_up_the_bbq_sauce_and_place_it_in_the_basket_seed2_step0007_003639
test
LC8_TASK_SEMANTIC_BINDING
multi_choice
LB_C8_MC_TASK_TARGETS_08
LB_C8_MC_TASK_TARGETS_08_v01
[ "A", "B", "C", "D" ]
[ "basket", "ketchup", "bbq sauce", "salad dressing" ]
{"A":"basket","B":"ketchup","C":"bbq sauce","D":"salad dressing"}
list
Set Exact Match + F1
set_matching
libero_gt_objective_scorer.score_one
Gold and prediction are sets of option letters; report exact match, precision, recall, and F1.
true
Bind the natural object names in the task text to the current object category set.
[ "task_language", "ground_truth.objects" ]
{"answer_evidence":{"distractors":["ketchup","salad dressing"],"target_categories":["basket","bbq sauce"],"task_language":"pick up the bbq sauce and place it in the basket"},"gt_lineage":{"difficulty":"hard","distractor_strategy":"same-scene non-target objects as distractors","gt_lineage_depth":3,"involved_gt_edges":["...
task type: move the yellow and white mug in front of the white mug
[ "A. place/move object", "B. open object", "C. turn on the switch or device", "D. close object", "E. other operations", "F. pick up object" ]
A
single_choice
task semantics bound to goals
Multiple Choice Question
medium
LIBERO-OBJ-000292
libero_90_KITCHEN_SCENE6_put_the_yellow_and_white_mug_to_the_front_of_the_white_mug_seed3_step0006_001438
test
LC8_TASK_SEMANTIC_BINDING
single_choice
LB_C8_SC_TASK_ACTION_01
LB_C8_SC_TASK_ACTION_01_v04
[ "A", "B", "C", "D", "E", "F" ]
[ "place/move object", "open object", "turn on the switch or device", "close object", "other operations", "pick up object" ]
{"A":"place/move object","B":"open object","C":"turn on the switch or device","D":"close object","E":"other operations","F":"pick up object"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The task language maps to action families through fixed verb rules.
[ "task_language" ]
{"answer_evidence":{"action_family":"place/move object","task_language":"put the yellow and white mug to the front of the white mug"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"other action families as semantic distractors","gt_lineage_depth":1,"involved_gt_edges":["action_family_parse"],"involved_gt_no...
Please identify which annotated panel corresponds to the eye-in-hand viewpoint.
[ "A. Panel A", "B. Panel B" ]
A
multi_view_choice
Dual-view Consistency and Camera Recognition
Multi-view Consistency Multiple Choice
simple
LIBERO-OBJ-000013
libero_90_LIVING_ROOM_SCENE1_pick_up_the_alphabet_soup_and_put_it_in_the_basket_seed0_step0004_001796
test
LC9_MULTI_VIEW_CONSISTENCY
multi_view_choice
LB_C9_MV_CAMERA_08
LB_C9_MV_CAMERA_08_v03
[ "A", "B" ]
[ "Panel A", "Panel B" ]
{"A":"Panel A","B":"Panel B"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Two synchronized RGB camera sources are randomly arranged, and the answer is uniquely determined by the camera key.
[ "images.agentview_image", "images.robot0_eye_in_hand_image" ]
{"answer_evidence":{"panel_sources":{"A":"robot0_eye_in_hand_image","B":"agentview_image"}},"gt_lineage":{"difficulty":"simple","distractor_strategy":"camera order randomized","gt_lineage_depth":1,"involved_gt_edges":["camera_source_mapping"],"involved_gt_nodes":["image:robot0_eye_in_hand_image","image:agentview_image"...
Select from the options a category of object present in the current scene.
[ "A. moka pot", "B. new salad dressing", "C. alphabet soup", "D. black book" ]
D
single_choice
visible object list and category recognition
Multiple Choice Question
simple
LIBERO-OBJ-000075
libero_90_STUDY_SCENE4_pick_up_the_book_on_the_right_and_place_it_under_the_cabinet_shelf_seed1_step0006_003182
test
LC1_OBJECT_INVENTORY
single_choice
LB_C1_SC_OBJECT_PRESENT_10
LB_C1_SC_OBJECT_PRESENT_10_v02
[ "A", "B", "C", "D" ]
[ "moka pot", "new salad dressing", "alphabet soup", "black book" ]
{"A":"moka pot","B":"new salad dressing","C":"alphabet soup","D":"black book"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
One correct category comes from the current object set, while the rest are categories that do not appear in that scene from the same dataset.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"answer_category":"black book","distractors":["moka pot","new salad dressing","alphabet soup"]},"gt_lineage":{"difficulty":"simple","distractor_strategy":"absent categories from global LIBERO object pool","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":["object:black_book_1"]},"invol...
According to the scene identifier, is the alphabet soup in the scene?
[ "A. is", "B. No" ]
B
yes_no
visible object list and category recognition
Binary judgment question
simple
LIBERO-OBJ-000382
libero_90_KITCHEN_SCENE2_open_the_top_drawer_of_the_cabinet_seed0_step0003_000675
test
LC1_OBJECT_INVENTORY
yes_no
LB_C1_YN_OBJECT_PRESENT_09
LB_C1_YN_OBJECT_PRESENT_09_v05
[ "A", "B" ]
[ "is", "No" ]
{"A":"is","B":"No"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
Existence of a single category is directly inferred from the object's state.
[ "ground_truth.objects.*.position" ]
{"answer_evidence":{"present":false,"queried_category":"alphabet soup"},"gt_lineage":{"difficulty":"simple","distractor_strategy":"same-workspace absent category distractor","gt_lineage_depth":1,"involved_gt_edges":[],"involved_gt_nodes":[]},"involved_objects":[],"provenance":{"episode_idx":0,"gt_fields":["ground_truth...
Which range is closest to the number of different object categories?
[ "A. 0-2 classes", "B. 6-8 categories", "C. 9 categories and above", "D. 3-5 classes" ]
B
interval_choice
quantity range and set judgment
Interval Selection Question
medium
LIBERO-OBJ-000042
libero_object_pick_up_the_butter_and_place_it_in_the_basket_seed1_step0004_003724
test
LC2_COUNT_AND_SET
interval_choice
LB_C2_INT_CATEGORY_COUNT_08
LB_C2_INT_CATEGORY_COUNT_08_v04
[ "A", "B", "C", "D" ]
[ "0-2 classes", "6-8 categories", "9 categories and above", "3-5 classes" ]
{"A":"0-2 classes","B":"6-8 categories","C":"9 categories and above","D":"3-5 classes"}
scalar
Accuracy
choice_accuracy
libero_gt_objective_scorer.score_one
Normalize prediction to an option letter and compare with the gold option letter.
true
The number of category classes after deduplication maps to a fixed range.
[ "ground_truth.objects" ]
{"answer_evidence":{"category_count":7,"interval":"6-8 categories"},"gt_lineage":{"difficulty":"medium","distractor_strategy":"fixed category-count interval bins","gt_lineage_depth":1,"involved_gt_edges":["unique_category_count"],"involved_gt_nodes":["category:basket","category:bbq sauce","category:butter","category:ch...
End of preview. Expand in Data Studio

BenchForge LIBERO

Object and spatial reasoning in simulated robotic manipulation.

The test split contains 10,000 English questions. The English edition preserves the original question IDs, option labels, answers, and scoring definitions. The main image and all referenced auxiliary images are embedded in the Parquet files as their original image bytes.

  • image: Main image supplied to the question.
  • auxiliary_images: Ordered list of additional images supplied with the question. An empty list means the task has no additional images.
  • question, choices, and answer: English question and answer choices, with the original option labels in the scoring answer.
  • data/test-00000-of-00001.jsonl: English text records with their original relative image paths for users of the existing benchmark layout.

Load the English test split using Hugging Face Datasets:

from datasets import load_dataset

dataset = load_dataset("jby20180901/BenchForge-libero", split="test")

Use the Parquet edition to access images directly. For the JSONL edition, use the repository’s assets/ directory or extract assets.tar at the repository root if the dataset ships with an image archive.

Downloads last month
53