Update oversight reward curve PNG

Update commander reward curve PNG

Use relative paths for reward curve images

Sync gitignore and latest README/BLOG from main

Use GitHub raw URLs for reward curve images

Add blog post, update README with Colab link and training results

docs: update plan.md with training results, inference benchmarks, new files

demo: exercise citation pathway + deterministic artifact + CI + walkthrough

Fix inference metadata field, add benchmark results, update README file layout links

Add HF Space link, results table, and reward curve to README

Add training results: reward curves, results doc, updated dashboard

Fix oversight reward: expose oversight_reward on IncidentObservation

Save full training logs to JSON + fix oversight obs.metadata crash

Reduce to 4 gens/200 tokens for ~60s/step on T4 (120 steps ≈ 2h)

Add reward and done fields to IncidentObservation

Fix env crash: remove obs.metadata assignment missing from IncidentObservation

Print env crash traceback in reward fn to diagnose -0.3 path

Add reward diagnostic logging to identify why rewards are flat

Tune GRPO for T4 speed: 4 gens, 256 tokens, temp 1.2

Add training dashboard with per-model reward curves and metrics

Fix GRPO rollout collapse: temperature 0.7→1.0, generations 4→8, completion length 300→512

Add dashboard.html and fix gitignore for runs/

Citadel v2.0.0 — Multi-Agent AI Defense Council