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

chore: refresh fixture summary artifact

refactor: replace unsloth with plain transformers+peft for GRPO training

feat: add verifier-native reward v1

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

feat: add real-time stellarator optimization demo animation

docs: align training workflow and plan ssot

refactor: align colab notebook with shared llm helpers

feat: make llm training workflow low-fidelity only

feat: add model-driven llm reward evaluation

fix: align notebook evaluation with grpo rewards

fix: align llm_agent auto-submit and reward handling with notebook

fix: robust JSON array extraction and notebook GRPO fixes

feat: add reward and verifier monitoring telemetry

feat: add HF Space landing page, revert to 2 reward functions

docs: clarify notebook surfaces and OpenEnv guidance

fix: restore ppo smoke early termination

fix: harden ppo smoke action handling

feat: polish notebook and README for hackathon submission

feat: add llm rollout contract and simplify ppo smoke

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

fix: quote pip version specs in Dockerfile, add cpu-upgrade

feat: refresh heuristic baseline and sync docs

feat: add HF Space deployment + GRPO training notebook

update docs

feat: add low-fi PPO smoke workflow

docs: clarify measured sweep caveats

docs: archive submission plan with SSOT-aligned fixes

fix: resolve high-fi submit blocker by switching to from_boundary_resolution preset

feat: add high-fidelity validation evidence

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

docs: add replay playtest report with reward branch coverage

fix: align client with openenv payload

docs: codify multifidelity training policy

feat: add replay playtest and tighten fail-fast validation

docs: align hybrid validation gates

docs: adopt hybrid fail-fast validation order

docs: require trained policy evidence

docs: clarify measured sweep provenance

docs: add measured sweep report and artifacts

docs: polish verifier reward review

feat: add initial p1 validation artifacts

docs: add verifier reward review

docs: simplify and archive planning surfaces

refactor: align p1 runtime contract and baseline reporting

fix: preserve recovery signal after VMEC failure

docs: clarify repaired env runtime status

docs: align repaired 4-knob environment docs

fix: align submit scoring with fidelity

feat: implement repaired p1 parameterization