SII-GAIR applies agentic self-improvement to full 30B-parameter pretraining run

SII-GAIR and Xinxu AI applied a recursive self-improvement (RSI) loop to a full frontier-scale pretraining run, training a 30B-parameter model (3B active) on 134B agentic tokens, and plan to open-source the weights.

2026-11-05 ~ 2026-11-05 · 2 related posts

• SII-GAIR applies RSI to full frontier-scale pretraining: 30B model trained on 134B agentic tokens — AdinaYakup · 2026-11-05
• SII-GAIR runs RSI on a full frontier-scale pre-training: 30B/A3B with 134B agentic tokens — AdinaYakup · 2026-11-05