October 4, 2026 · The Decoder
Google's RRSI method stops self-improving agents from memorizing tests
Self-improving AI agents tend to memorize their test tasks, so gains shrink on new ones. Google researchers' RRSI method regularizes this effect. It lifts scores on unseen benchmarks by up to 4.7 points and uses about 30 percent fewer tokens than an unregularized version.
Why it matters: Overfitting to the evaluation set is the central credibility problem for recursive self-improvement: gains that vanish on unseen tasks are not real capability. Paired with DeepMind's Dream-RSI work on cutting search iterations, this suggests labs are now focused on making self-improvement loops both cheaper and honest, which matters as agents increasingly tune themselves.
Related updates
- Transformer compresses 1GB of Wikipedia to 105MB losslesslyNov 21
- ByteDance's open Axion 1.0 agent rivals top models with three training phasesNov 21
- Stanford posts all-human Riemann hypothesis proof hours after Anthropic's Claude claimNov 21
- Thinking Machines releases open Inkling 2 Mini, a looped model for laptopsNov 21