Reproduction record

From retained records to published claims

This page identifies the artifacts and commands behind the numerical claims on the project site. It documents the completed work and does not authorize new training runs or robot trials.

Evidence boundary

What can be reproduced

Values derived from data are recomputed from retained CSV, Parquet, JSON, and annotation files. Configuration values are checked against policy code, resolved commands, checkpoint configuration, and evaluation manifests. Literature claims are checked against same-key source files in the reference archive.

Dataset

Collection and calibration lineage

ComponentAuthoritative recordVerified property
Peg-in-hole datasetexperiments/datasets/peg-tight-vertical-100/Episode count, frame count, timestamps, camera keys, states, and actions
Controlled viewsact-v/meta/ and act-m/meta/Identical episode and frame identities with only the state view changed
Leader calibrationcalibration/so101_leader.jsonJoint order, servo identifiers, ranges, homing offsets, and digest
Follower calibrationcalibration/so101_follower.jsonThe same fields and digest, fixed from collection through evaluation
Models and trials

Identity before interpretation

Resolved command records preserve dataset view, seed, update budget, policy family, resolution, and chunk or horizon settings. Evaluation manifests bind deployed labels to checkpoint hashes and policy-code revisions. Trial rows preserve order, outcome, failure phase, and session context.

Verification

Deterministic checks

Manuscript auditpython3 research/writing/audit_final_manuscript.py
Dataset verificationpython3 research/writing/verify_chapter5_dataset.py
Evaluation verificationpython3 research/writing/verify_chapter7_evaluation.py

Large recordings and checkpoint tensors remain in the project archive and are not embedded in this website.