license: apache-2.0
pretty_name: DROID QC eval set
task_categories:
- video-classification
- image-classification
tags:
- robotics
- teleoperation
- data-quality
- vlm-as-a-judge
- calibration
- droid
- lerobot
size_categories:
- n<1K
source_datasets:
- lerobot/droid_1.0.1
configs:
- config_name: default
data_files:
- split: test
path: data/test-*.parquet
DROID QC eval set
200 teleoperation episodes from DROID 1.0.1
(100 labelled successful, 100 failed by the people who collected them), the frame grid a
vision-language judge sees for each, and the verdicts of claude-opus-5-5. Built with
teleop-qc, which measures how far a VLM judge of robot task success can
be trusted before it replaces human review in a data pipeline.
What is in a row
| column | meaning |
|---|---|
episode_index |
episode in lerobot/droid_1.0.1 |
grid |
the image the judge sees (968x1100 JPEG): 3 cameras in columns, 6 moments in rows, first at the top |
human_success |
DROID's success label, set by the operator who collected the episode |
instruction |
DROID's language instruction, "" when missing. Never shown to the judge (see below) |
frames, seconds |
episode length at 15 fps |
videos |
where each camera stream lives in the DROID video files (JSON) |
judge_* |
the verdict: success, confidence (0.5 to 1), p_success, failure mode, evidence, attempts, tokens |
Columns of the grid, left to right: exterior camera 1, exterior camera 2, wrist camera. judge_p_success is the
confidence for a success verdict and 1 - confidence for a failure. The judge abstains instead of guessing when
its answer fails a schema and consistency check twice: abstentions have null judge_success and
judge_p_success.
How the episodes were drawn
100 successful and 100 failed episodes, drawn with seed 0 from the first metadata file of DROID 1.0.1
(14,153 episodes). The set is frozen as evalset/droid_evalset_v1.jsonl; a change makes a new version.
Why the judge never sees the instruction. In that metadata file, 94.6% of the episodes labelled successful have a language instruction, against 0.3% of the failed ones: the instructions were crowdsourced after collection, mostly for successful episodes. A judge shown the instruction could read the label off its presence.
Results of claude-opus-5-5
Command: teleop-qc judge --evalset evals/droid_evalset_v1.jsonl --out runs/claude-opus-5-5.jsonl (effort medium,
structured outputs), then teleop-qc report.
| metric | value |
|---|---|
| episodes judged | 200 of 200 (0 abstentions) |
| balanced accuracy at p = 0.5 | 0.755 |
| human successes judged successful | 0.830 |
| human failures judged failed | 0.680 |
| Cohen's kappa | 0.510 |
| AUROC (95% bootstrap interval) | 0.811 (0.747 to 0.868) |
| Brier score | 0.179 |
| ECE (10 uniform bins) | 0.064 |
| ECE after cross-fitted isotonic recalibration | 0.051 |
| share accepted without review at 95% precision | 0.000 |
| share rejected without review at 90% precision | 0.130 |
| cost of the run | $2.92 |
| cost per episode | $0.0146 |
| cost per hour of episode | $2.89 |
Files
data/test-*.parquet: the table above, grids included.evalset/droid_evalset_v1.jsonl: the frozen eval set, as read byteleop-qc judge.verdicts/claude-opus-5-5.jsonl: the judge's raw output, one line per episode.judge/system_prompt.txt,judge/answer_schema.json: what the judge is told and the JSON schema its answer must follow.
License and citation
Apache-2.0, like DROID. The frames come from DROID:
@article{khazatsky2024droid,
title = {DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset},
author = {Khazatsky, Alexander and Pertsch, Karl and others},
journal = {arXiv preprint arXiv:2403.12945},
year = {2024}
}