MarcelTorne/put_cube_in_bowl_preferences

put_cube_in_bowl — pairwise preferences on a Franka Panda

Real-robot trajectories for "put the cube in the bowl" with human pairwise preference labels on multiple judgment axes. Built for reward-model / preference-learning research: every label is a comparison of two trajectories on one named axis, not a scalar score.

The trajectory data is a standard LeRobot v2.1 dataset, so it also loads directly as an imitation-learning dataset.

Contents

Episodes	540
Frames	65,083 (~1.2 h at 15 fps)
Preference pairs	808
Preference labels (pair x axis)	2348
Judgment axes	6
Distinct instructions	15
Cameras	agent_view (third-person), wrist — both 224x224
Robot	Franka Panda, 7-DoF joint control + gripper

Episodes by kind

kind	episodes	frames
rollout	420	57199
demo	120	7884

demo episodes are human teleoperated demonstrations. rollout episodes are policy rollouts recorded in preference-collection sessions; those are the ones that vary in quality, which is what makes the comparisons informative.

Language instructions

This task was collected under several phrasings, and they are not interchangeable -- most name which bowl is the target. Each episode keeps its own instruction as its LeRobot task string.

instruction	episodes
put the cube in the bowl	120
put the cube in the blue bowl	70
put the cube in the yellow bowl	66
put the cube in the orange bowl	64
Pick up the orange cube and place it in the blue bowl	38
Pick up the orange cube and place it in the yellow bowl	36
Pick up the orange cube and place it in the orange bowl	26
put the orange cube in the yellow bowl	26
put the orange cube in the blue bowl	24
put the orange cube in the orange bowl	20
put the orange cube in the bowl	16
put orange cube in the blue bowl	10
put orange cube in the bowl	10
put orange cube in the yellow bowl	8
put orange cube in the orange bowl	6

Preference labels

preferences/pairs.parquet — one row per (pair, axis):

column	meaning
pair_id	unique id of the comparison
episode_index_a, episode_index_b	index into this LeRobot dataset
episode_id_a, episode_id_b	original source trajectory id
axis	the judgment axis being compared
winner	A, B, or Equal
source	in_session (the two rollouts of one session) or cross (arbitrary pair)
source_dir, source_file	provenance of the annotation
annotator	which annotator's label set this row comes from
axis_set	fixed (curated rubric reused across sessions) or freeform (annotator-invented axis names)
overall_score_a/b	1–4 Likert quality rating, where the annotator gave one (else null)
succeeded_a/b	task-success flag recorded at collection time, where available (else null)
instruction	instruction string logged with the comparison

preferences/pairs.jsonl is the same data with the axes nested per pair. preferences/episodes.parquet maps episode_index to the source trajectory.

Winner distribution: B 871, A 829, Equal 648.

Axes

axis	labels
Overall quality	808
Fast	308
Smooth	308
Blue bowl	308
Orange bowl	308
Yellow bowl	308

Pairs by source

source	directory	axis set	pairs
cross	abhijnya/cross_preferences_pick_and_place	fixed	250
cross	am208/cross_preferences_pick_and_place	fixed	348
in_session	am208/preferences_pick_and_place	fixed	210

Usage

from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("MarcelTorne/put_cube_in_bowl_preferences")
print(ds[0].keys())

# train a reward model on the pairwise labels
import pandas as pd
from huggingface_hub import hf_hub_download
prefs = pd.read_parquet(hf_hub_download("MarcelTorne/put_cube_in_bowl_preferences", "preferences/pairs.parquet",
                                        repo_type="dataset"))
decisive = prefs[prefs.winner != "Equal"]           # drop ties
overall  = decisive[decisive.axis.str.lower().str.contains("overall")]

Both cameras are dtype: video. observation.state is [joint_0..joint_6, gripper], observation.ee is [x, y, z, roll, pitch, yaw], and action is the commanded [joint_position(7), gripper_position(1)].

Caveats

• Frames are h264 (crf 16), not bit-exact. The source HDF5 held raw uint8 frames; re-encoding at the same 224x224 resolution shrinks the release ~100x at 41.8 dB PSNR (mean absolute error 1.1/255, 99th pct 7/255), measured on fold_pants cloth texture. Do not expect byte-level reproduction of the originals.
• Equal is common on some axes. Annotators used it freely; filter deliberately.
• freeform axis names are not a closed vocabulary. They were invented per session, so the same idea appears under several names (fast / speed, smooth / smoothness). Normalise before aggregating across sessions.
• A pair can be labelled by more than one annotator. The two annotators' cross-pair sets are near-disjoint, but not perfectly; group by pair_id if you need unique pairs.
• Instructions on rollout episodes sometimes carry reward conditioning (e.g. "put the cube in the bowl, fast: 1.0, ...") because the policy that produced them was reward-conditioned. The LeRobot task string is the plain instruction; the logged string is kept in preferences/episodes.parquet.
• No success labels on demo episodes — they are demonstrations, assumed good.

Related datasets

Same robot, same collection pipeline, same schema — four tasks released together:

• MarcelTorne/fold_pants_preferences — fold the shorts
• MarcelTorne/setup_table_preferences — set up the table
• MarcelTorne/put_cube_in_bowl_preferences (put the cube in the bowl) — this dataset
• MarcelTorne/plate_toast_preferences — put the toast in the plate

License

MIT.