All pull requests

Pull requests

• lance: Aria episode tables, loader, committer and pilot tools

  #716·RyanPCo opened Oct 14, 2026Contributor·

  ·

• history: learned null token for dropped and episode-start-padded history

  #715·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• mecka private: fixed-budget operator-scaling ladder generator

  #714·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• train: hardened WSD defined in-config, W&B experiment tag, mode=validate on EMA weights

  #713·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• val: exact-index metric frames, pad-masked metrics with thirds, per-loader eval RNG

  #712·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• refactor(transforms): separate rotation representation from action layout and frame

  #711·AnikethCheluva opened Oct 12, 2026Collaborator·

  ·
  1

• feat: add fixed-budget operator scaling experiments

  #710·AnikethCheluva opened Oct 12, 2026Collaborator·

  ·

• mecka private: mono converter, opscale configs and generators, IDs kept out of the repo

  #709·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• data/rdt: keypoint RDT recipe, pose-glitch masking, hardened staging

  #708·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• data: stage episodes to node-local disk; never read them from /workspace

  #707·RyanPCo opened Oct 12, 2026Contributor·

  ·
  1

• Evaluate adaptive Astra steering on a qualified native RoboCasa policy

  #706·rpuns opened Oct 6, 2026Contributor·

  ·

• Add reasoning-guided pi0.5 policy-learning pilot

  #705·rpuns opened Oct 3, 2026Contributor·

  ·

• Add Astra skill libraries over pi05 inputs and demo actions

  #704·rpuns opened Sep 30, 2026Contributor·

  ·

• ponytail(u1): drop bench/ and duplicate viz scripts, reuse gpu_orphans in the reaper

  #703·aidang3019 opened Sep 30, 2026Contributor·

  ·
  1

• eva: joint-space mode for the ABC YAM action space; robot image history

  #702·aidang3019 opened Sep 29, 2026Contributor·

  ·
  2

• pl_model: RDT-style weight EMA, validated in place of raw weights

  #701·RyanPCo opened Sep 28, 2026Contributor·4 tasks·

  ·
  1

• Evaluate Astra correction recipes and Codex-guided LIBERO interventions

  #700·rpuns opened Sep 27, 2026Contributor·

  ·

• Compare Astra text and visual representation steering

  #699·rpuns opened Sep 26, 2026Contributor·

  ·

• Evaluate Astra flow reversal and critique-driven noise policy learning

  #698·rpuns opened Sep 26, 2026Contributor·

  ·

• Test actual RGB perturbations with observed Astra feedback

  #697·rpuns opened Sep 25, 2026Contributor·

  ·

• Evaluate online Astra interpolation on frozen pi05

  #696·rpuns opened Sep 25, 2026Contributor·

  ·

• Evaluate iterative Astra noise, language, and vision interventions

  #695·rpuns opened Sep 25, 2026Contributor·

  ·

• Report paired pi0.5 OOD outcomes and Astra reversal failures

  #694·rpuns opened Sep 24, 2026Contributor·

  ·

• Run real Astra flow reversal and paired LIBERO-OOD jobs on OSMO

  #693·rpuns opened Sep 24, 2026Contributor·

  ·

• Add frozen pi0.5 flow reversal and reproducible LIBERO evaluation

  #692·rpuns opened Sep 24, 2026Contributor·

  ·