RSI-Anything: Unlocking Recursive Self-Improvement with World Models
Taoyong Cui, Pheng Ann Heng, Zhenfei Yin, Ling Yang and colleagues (PhAI Labs, CUHK, with Oxford, Stanford and PKU) present RSI-Anything, a recursive self-improvement framework that uses a predictive world model to choose actions and training data, then adapts the world model as the policy changes.
Ask this paper
World model as selector. A JEPA-Anything predictive world model scores candidate actions and learning data before they are spent on costly external feedback. Observed feedback then updates both the policy or workflow and the world model.
Retention checks. Component and compatibility checks decide whether an updated policy and world-model pair is kept, so a policy update that makes the world model unreliable is rejected.
Code repair. With the coder fixed, success rises from 26.3% with a static verifier to 29.2% with predictive selection and 31.2% after world-model adaptation, at comparable test usage (paired gain +2.83 points, 95% interval +0.17 to +5.50).
Planning and science. Laboratory-planning scores rise from 21.80% to 46.01% across five policy versions, and reaction optimization reaches 96.15% mean best yield against 82.16% for Bayesian optimization at the same 40-experiment budget.
GUI agents. Task completion rises while the average number of actions falls by 39.4%.
Abstract
Recursive self-improvement (RSI) aims to transform interaction experience into persistent improvements, yet external feedback is costly and world-model predictions may become unreliable as policies evolve. We introduce RSI-Anything, a domain-general framework built on JEPA-Anything that uses predictive world models to guide action and learning-data selection. Observed feedback drives policy or workflow improvement and conditional world-model adaptation, while component and compatibility checks govern the retention of updated policy-model pairs. Across diverse computational and scientific domains, RSI-Anything demonstrates substantial gains. In code repair, with the coder fixed, success increases from 26.3% with a static verifier to 29.2% through predictive selection and 31.2% after world-model adaptation, at comparable test usage. Graphical user interface (GUI) evaluations showed higher task completion, while average actions decrease by 39.4%. Across five successive policy versions, laboratory-planning scores more than double from 21.80% to 46.01%. In reaction optimization, RSI-Anything achieves 96.15% mean best observed yield versus 82.16% for Bayesian optimization under the same 40-experiment budget. An end-to-end immune-science application further connects workflow evolution and predictive modeling with therapeutic hypothesis generation. These results demonstrate how predictive world models and retained system updates enable recursive improvement across diverse environments.