Commit History

17 days ago

  1. Evaluation: fine-tune data scaling complete (16/16 points; water pretrained 10 / 30 added)

    Verified signature
    82e2305

18 days ago

  1. Link to the findings page

    Verified signature
    eb67dbc
  2. Fine-tune data scaling figures (10 of 12 points)

    Verified signature
    d3669d4
  3. Datasets: water flowers v3 complete (500 episodes)

    Verified signature
    9295134
  4. Datasets: water flowers v3 (realistic fixtures)

    Verified signature
    0d639d2
  5. Datasets section (HF dataset repo, per-episode downloads); pretraining-scaling figures (partial)

    Verified signature
    e7ec5ea
  6. Evaluation figures: data-v2 funnel readability

    Verified signature
    e72bae1
  7. Evaluation section for the data-v2 arms

    Verified signature
    4fc40a6
  8. Data v2 rollouts (A2 / A2′ / C2 + v1 baselines A / A′ / C / E′); 600 episodes

    Verified signature
    4fe45ee