Skip to page content
SAVRN
Search Plan a project

Dataset

b1k-task55-re-shelving-library-books-official-eval

by Jeremy jermo1120/b1k-task55-re-shelving-library-books-official-eval

Rollouts of the Pi0.5 fine-tune jermo1120/b1k-task55-re-shelving-library-books (step24999) on BEHAVIOR-1K task reshelvinglibrarybooks, following the 2026 challenge leaderboard protocol. Missing instances count as 0 in Q.

Rows—
Configurations—
Size273.2 MB
Licenseapache-2.0
AccessPublicly accessible
Monthly Downloads—

Dataset Card

By Jeremy, published under apache-2.0, revision c25ce1c07f10.

Official-protocol evaluation of jermo1120/b1k-task55-re-shelving-library-books

Rollouts of the Pi0.5 fine-tune jermo1120/b1k-task55-re-shelving-library-books (step_24999) on BEHAVIOR-1K task re_shelving_library_books, following the 2026 challenge leaderboard protocol.

Status: complete.

Result

checkpoint instances run successes Q (mean q_score over the 10 instances)
step_24999 10 / 10 0 0.067

Missing instances count as 0 in Q.

checkpoint instance_id success q_score steps norm. base / left / right distance
step_24999 301 No 0.333 14293 1.11 / 0.36 / 0.43
step_24999 302 No 0.000 14293 4.53 / 0.69 / 0.48
step_24999 303 No 0.000 14293 1.79 / 0.48 / 0.30
step_24999 304 No 0.000 14293 3.43 / 0.53 / 0.81
step_24999 305 No 0.000 14293 0.94 / 0.41 / 0.40
step_24999 306 No 0.000 14293 3.09 / 0.61 / 0.67
step_24999 307 No 0.000 14293 2.55 / 0.57 / 0.53
step_24999 308 No 0.000 14293 1.32 / 0.53 / 0.51
step_24999 309 No 0.333 14293 0.90 / 0.36 / 0.37
step_24999 310 No 0.000 14293 1.87 / 0.55 / 0.37

q_score gives partial credit: 1.0 for a full success, otherwise the fraction of goal predicates that were false at the start and true at the end of the episode. The task 55 goal is only "every book (3) is inside a bookcase", so per-instance scores are 0, 0.33, 0.67 or 1.0. Distances are normalized by the mean human demo (1.0 = same path length as a human).

Per-rollout metrics are in results/<checkpoint>/*.json and results/summary.csv, videos (head + both wrist cameras) in videos/<checkpoint>/, evaluator logs in logs/<checkpoint>/instance_*.log and the policy server log in logs/<checkpoint>/server.log.

Protocol

As specified in BEHAVIOR-1K docs/challenge/evaluation.md:

Instances public_test instance indices 0-9 (instance ids 301-310), the leaderboard-reporting half of the split
Rollouts 1 per instance, no reruns or selection
Episode limit default 1.5x mean human-demo length (14292 steps)
Videos --write-video
Evaluator BEHAVIOR-1K v3.9.3-post2 (python -m omnigibson.eval.eval), OmniGibson 3.9.3, Isaac Sim 5.1
Robot config omnigibson/eval/r1pro.yaml (robot robot_r1), default wrapper (224x224 RGB + proprio)
Policy server wensi-ai/openpi@behavior + the model repo's changes.patch + scripts/robot_config_eval2026.patch; scripts/b1k/serve_b1k.py --robot b1k/R1Pro_eval2026
Hardware 1x RTX 5090 (simulator and policy on the same GPU)

Command: OUT_ROOT=out_official INDICES="0 1 2 3 4 5 6 7 8 9" scripts/run_eval.sh 24999.

Checkpoint selection

step_24999 was chosen from a sweep of 9 checkpoints (5k to 25k) on the other half of the split (instances 311-315, indices 10-14), where it was the only checkpoint with nonzero q_score. The sweep is in jermo1120/b1k-task55-re-shelving-library-books-eval.

Details

Repository
jermo1120/b1k-task55-re-shelving-library-books-official-eval
Publisher
Jeremy
Task category
Not stated by the source
Tags
robotics, behavior-1k, evaluation
Size category
Not stated by the source
Languages
Not stated by the source
Revision
c25ce1c07f1001dc5fafe710db12c54f7e7e7533
Last updated
2026-10-13

Files

37 files, 273.2 MB in total.

Data11 files · 6.2 KB
Documentation1 file · 3.4 KB
Other24 files · 273.1 MB
Repository1 file · 2.5 KB
Every file
FileTypeSizeSHA-256
results/step_24999/re_shelving_library_books_301_0.jsonData561 B—
results/step_24999/re_shelving_library_books_302_0.jsonData543 B—
results/step_24999/re_shelving_library_books_303_0.jsonData544 B—
results/step_24999/re_shelving_library_books_304_0.jsonData544 B—
results/step_24999/re_shelving_library_books_305_0.jsonData541 B—
results/step_24999/re_shelving_library_books_306_0.jsonData544 B—
results/step_24999/re_shelving_library_books_307_0.jsonData543 B—
results/step_24999/re_shelving_library_books_308_0.jsonData543 B—
results/step_24999/re_shelving_library_books_309_0.jsonData557 B—
results/step_24999/re_shelving_library_books_310_0.jsonData543 B—
results/summary.csvData730 B—
README.mdDocumentation3.4 KB—
logs/step_24999/instance_301.logOther56.6 KB—
logs/step_24999/instance_302.logOther55.8 KB—
logs/step_24999/instance_303.logOther55.8 KB—
logs/step_24999/instance_304.logOther55.8 KB—
logs/step_24999/instance_305.logOther56.1 KB—
logs/step_24999/instance_306.logOther57.4 KB—
logs/step_24999/instance_307.logOther56.1 KB—
logs/step_24999/instance_308.logOther56.1 KB—
logs/step_24999/instance_309.logOther56.9 KB—
logs/step_24999/instance_310.logOther56.1 KB—
logs/step_24999/server.logOther12.7 KB—
scripts/build_official.pyOther5.1 KB—
scripts/robot_config_eval2026.patchOther1.1 KB—
scripts/run_eval.shOther1.7 KB—
videos/step_24999/re_shelving_library_books_301_0.mp4Other27.3 MBafea22511d93
videos/step_24999/re_shelving_library_books_302_0.mp4Other22.9 MBef0e7ce24d11
videos/step_24999/re_shelving_library_books_303_0.mp4Other27.6 MB63dbbb94ec7d
videos/step_24999/re_shelving_library_books_304_0.mp4Other28.6 MBaa8583d764ff
videos/step_24999/re_shelving_library_books_305_0.mp4Other27.4 MB0c5230ce1f3c
videos/step_24999/re_shelving_library_books_306_0.mp4Other24.2 MB5182f304b7b5
videos/step_24999/re_shelving_library_books_307_0.mp4Other27.6 MB400f81ab22d7
videos/step_24999/re_shelving_library_books_308_0.mp4Other28.1 MB229eb5c8be06
videos/step_24999/re_shelving_library_books_309_0.mp4Other29.1 MB74989f9e02d8
videos/step_24999/re_shelving_library_books_310_0.mp4Other29.6 MB382e1848f819
.gitattributesRepository2.5 KB—

License and Download

License
apache-2.0
Access
No access gate
Download from Jeremy

Released by Jeremy through its official repository on Hugging Face. Read the license.