Google DeepMind's SIM method lets AI self-improvement experience transfer across models
智东西 · wechat · 2026-11-19
Researchers from Google Cloud AI, Google DeepMind and Penn State published a paper proposing Self-Improvement Memory (SIM), a method enabling continual recursive self-improvement (CRSI). Unlike prior RSI work that only keeps modified harness code, SIM records each change's rationale, expected and actual outcomes—like a searchable experiment log—so optimization experience survives model and task changes.
In 10 controlled coding experiments, SIM-enabled runs won 9 of 10, averaging 2.36 points higher (p=0.003). When a harness optimized for Claude was migrated to Gemini, SIM-enabled optimization topped all four domains, reaching 84.75 on office tasks versus 73.42 without SIM. But inherited experience isn't always a win—it trailed by 0.8 points on Gemini 4 Argon.
Notably, model weights stayed frozen throughout; the AI improves the external harness, not the model itself. First author Xin Yu, a Nankai University statistics graduate now at Penn State, completed the work during a Google internship.
More from AGI Musings
- Sam Altman Declares AGI Has Arrived, Says OpenAI Limits AI Self-Improvement for Safety — FinanceYF5 · 2026-11-19
- Joshua Saxe: AI policy progress needs empirical capacity-building, not endless debate — joshua_saxe · 2026-11-19
- Is prompting your way to a prize dishonest? The AI circle debates fairness in competitions — TuhinChakr · 2026-11-19
- Noah Smith: OpenAI is a research lab posing as a profit machine; Anthropic is the opposite — jdjohnson · 2026-11-19
- 'Wet lab is the new neocloud': bio lab capacity emerges as AI's next scarce resource — yawnxyz · 2026-11-19
- Kevin Bass: AI's biggest risk isn't becoming a god, but people who want god-like control over it — kevinnbass · 2026-11-19