RSIArena: 8 Research Agents Self-Study Post-Training for 144 Hours
RSIArena, with Stanford, Notre Dame, UW and Scale AI, is running 8 research agents on the same 30B base model for 144 hours on 64 RTX 6000 GPUs, livestreaming the experiment to test how much post-training research frontier models can do independently.
2026-09-30 ~ 2026-09-30 · 2 related posts
- 8 research agents self-train a 30B model for 144 hours in RSIArena livestream experiment — my_cat_can_code · 2026-09-30
- RSIArena pits 8 research agents on one 30B model with 64 RTX 6000 GPUs for 144 hours — my_cat_can_code · 2026-09-30