Fetching metadata from the HF Docker repository...
-
Initial OpenENV hackathon submission
ThejasRao +1: Claude Sonnet 4.6 Multi-step content moderation RL environment with:
- 24 deterministic seed-based scenarios (8 per difficulty, 4 geos)
- Dense reward + 4-component grader
- Hybrid LLM agent (deterministic investigation + LLM classify/decide)
- Rule-based baseline (easy=0.93, medium=1.00, hard=0.50)
- FastAPI on port 7860 (HF Spaces compatible)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>