Commit History

  1. feat: reward verifier alignment, notebook hardening, model name fix⁠

    CreativeEngineer +1: Claude Opus 4.6
    cdc237b
  2. chore: refresh fixture summary artifact

    1ddf57a
  3. refactor: replace unsloth with plain transformers+peft for GRPO training⁠

    CreativeEngineer +1: Claude Opus 4.6
    3313e24
  4. feat: add verifier-native reward v1

    2fccde8
  5. feat: upgrade notebook to Qwen3.5-4B with H100 hyperparams⁠

    CreativeEngineer +1: Claude Opus 4.6
    2cb6617
  6. feat: add real-time stellarator optimization demo animation⁠

    CreativeEngineer +1: Claude Opus 4.6
    3b185f9
  7. docs: align training workflow and plan ssot

    e8e5af5
  8. refactor: align colab notebook with shared llm helpers

    ddcb837
  9. feat: make llm training workflow low-fidelity only

    9c3599b
  10. feat: add model-driven llm reward evaluation

    ede4c5c
  11. fix: align notebook evaluation with grpo rewards

    8254ade
  12. fix: align llm_agent auto-submit and reward handling with notebook⁠

    CreativeEngineer +1: Claude Opus 4.6
    5f2da5f
  13. fix: robust JSON array extraction and notebook GRPO fixes⁠

    CreativeEngineer +1: Claude Opus 4.6
    e826e11
  14. feat: add reward and verifier monitoring telemetry

    5e0e606
  15. feat: add HF Space landing page, revert to 2 reward functions⁠

    CreativeEngineer +1: Claude Opus 4.6
    8d69cc6
  16. docs: clarify notebook surfaces and OpenEnv guidance

    2348d3e
  17. fix: restore ppo smoke early termination

    9827b11
  18. fix: harden ppo smoke action handling

    40011e5
  19. feat: polish notebook and README for hackathon submission⁠

    CreativeEngineer +1: Claude Opus 4.6
    c647aa0
  20. feat: add llm rollout contract and simplify ppo smoke

    ebd0ff3
  21. fix: add cmake, ninja, netcdf build deps for booz-xform⁠

    CreativeEngineer +1: Claude Opus 4.6
    729c711
  22. fix: quote pip version specs in Dockerfile, add cpu-upgrade⁠

    CreativeEngineer +1: Claude Opus 4.6
    cd1dff0
  23. feat: refresh heuristic baseline and sync docs

    f238af4
  24. feat: add HF Space deployment + GRPO training notebook⁠

    CreativeEngineer +1: Claude Opus 4.6
    3bfd80a
  25. update docs

    2dba2cf
  26. feat: add low-fi PPO smoke workflow

    97fc141
  27. docs: clarify measured sweep caveats

    0908e18
  28. docs: archive submission plan with SSOT-aligned fixes⁠

    CreativeEngineer +1: Claude Opus 4.6
    25f0cfa
  29. fix: resolve high-fi submit blocker by switching to from_boundary_resolution preset⁠

    CreativeEngineer +1: Claude Opus 4.6
    9d7dc15
  30. feat: add high-fidelity validation evidence

    c3a24db
  31. docs: add 14-hour submission plan with time allocation and cut priorities⁠

    CreativeEngineer +1: Claude Opus 4.6
    a02ffad
  32. docs: add replay playtest report with reward branch coverage⁠

    CreativeEngineer +1: Claude Opus 4.6
    bca349b
  33. fix: align client with openenv payload

    dc2e64b
  34. docs: codify multifidelity training policy

    513a2e2
  35. feat: add replay playtest and tighten fail-fast validation

    8bf0155
  36. docs: align hybrid validation gates

    567ff67
  37. docs: adopt hybrid fail-fast validation order

    1c1f314
  38. docs: require trained policy evidence

    27d58b3
  39. docs: clarify measured sweep provenance

    714d655
  40. docs: add measured sweep report and artifacts⁠

    CreativeEngineer +1: Claude Opus 4.6
    e8134ed
  41. docs: polish verifier reward review

    80ce5e5
  42. feat: add initial p1 validation artifacts

    92c59c2
  43. docs: add verifier reward review

    f09f601
  44. docs: simplify and archive planning surfaces

    e815b38
  45. refactor: align p1 runtime contract and baseline reporting

    6deaccc
  46. fix: preserve recovery signal after VMEC failure

    eb446cf
  47. docs: clarify repaired env runtime status

    5271cce
  48. docs: align repaired 4-knob environment docs

    3270c54
  49. fix: align submit scoring with fidelity

    88d9b78
  50. feat: implement repaired p1 parameterization

    fe3a41d