RSI Arena Stage 1 ends: grok 4.7's agent-trained model lands top two despite early cutoff
my_cat_can_code · x · 2026-10-12
Stage 1 of RSI Arena — where AI agents train models judged by human evaluation — has concluded. grok 4.7 ran out of API credit 41 hours early with nearly 70% of its GPU budget unused, yet its model still made the top two. Collaborators include Hugging Face, Scale AI, Notre Dame, UW and Stanford. Stage 2 is now live, adding human feedback to the training loop, with live streams, arena results and daily reports available.
More from Research
- Lilian Weng publishes 25-minute deep dive on the math of scaling laws — irinarish · 2026-10-12
- Professor recruiting PhD students at intersection of computational mechanics and computer vision — anand_bhattad · 2026-10-12
- JHU's FracGen teaches video models physics-accurate stretching and tearing from a single image — anand_bhattad · 2026-10-12
- EgoLAP paper adds motion-level language reasoning to VLA pretraining, lifting real-world robot task progress to 80.1% — Majumdar_Ani · 2026-10-12
- Sherpa RL framework trains LLM teachers by directly maximizing simulated student learning outcomes — chrmanning · 2026-10-12
- openjev, an open-source hallucination detection and reranking tool, trends on Hugging Face — AlexWortega · 2026-10-12