Spaces:
Sleeping
Sleeping
Commit History
-
Update oversight reward curve PNG
Verified signaturef74b27a -
Update commander reward curve PNG
Verified signatureed937c6 -
Add blog post, update README with Colab link and training results
e9a2950 -
docs: update plan.md with training results, inference benchmarks, new files
033c53a -
Add training results: reward curves, results doc, updated dashboard
- docs/results.md: full hackathon-aligned results document covering environment design, reward functions, curriculum, before/after metrics - docs/results/: commander + oversight reward curve PNG + JSON - runs/training_dashboard.html: updated with fixed oversight run data Commander: -0.326 → +0.082 avg (+0.41 improvement) Oversight: -0.145 → +0.134 avg (+0.28 improvement) Co-Authored-By: Shaurya Verma <shaurya.verma@scaler.com>bf77795
-
Citadel v2.0.0 — Multi-Agent AI Defense Council
Two-LLM council (Commander + Oversight) for enterprise SOC incident response. Covers Themes 1 (Multi-Agent / Fleet AI), 3.1 (Professional / Scaler AI Labs), 4 (Self-Improvement via adversary curriculum + shared playbook), and 5 (Wild Card — bidirectional trust dynamics).f0609d2