RSIArena livestreams 8 research agents post-training the same 30B model on 64 Blackwell GPUs
my_cat_can_code · x · 2026-09-30
RSIArena, run with Stanford, Notre Dame, UW, and Scale AI, is testing how much post-training research frontier models can do autonomously.
- 8 research agents, all starting from the same 30B base model
- 144 hours on a shared cluster of 64 RTX PRO 6000 Blackwell GPUs
- Each agent gets 1,000 GPU-hours to pick its data, write training code, and run experiments
- Submitted models are frozen and independently evaluated — "models training models," live
The team will host a livestream event and booth #107 at COLM 2026, and is seeking feedback from anyone working on post-training, agents, or evaluation.
More from coding & agent
- Generic API-catalog MCP server vs one tool per job: a context-bloat experiment — MandeepKaur9407 · 2026-11-16
- Berkeley paper: LLMs answer with stale values even when the update is in context — ombre-ai · 2026-11-16
- Effective HTML: 4.2k-star agent skills turn wireframes into self-contained interactive prototypes — tom_doerr · 2026-11-16
- skills-gate: one gate skill fronts 148 agent skills at just 389 tokens — Distinct-Pie2389 · 2026-11-16
- Open-weight Karpo-Decision-31B tops Decision Index at 73.9, one forward pass per answer — CodeByPoonam · 2026-11-16
- shadcn boosts @uisdk: one-shot UI rebuild that's instantly E2E-testable by AI — shadcn · 2026-11-16