Recursa: Toward Recursively Self-Improving Foundation Model Systems
Erkang Zhu, Jingren Zhou and colleagues at Alibaba Research present Recursa, a system that adapts to a new domain by learning from its own task experience, storing skills, instructions and memories in a knowledge repository that parallel learners refine without a human expert or stronger teacher.
Ask this paper
Repository. Recursa stores domain skills, instructions on when to apply them, memories of past tasks and reflections on what to improve.
Two phases. Bootstrapping builds the repository from domain materials. Recursive self-improvement then runs tasks with parallel learners, synthesizes local fixes, validates them and merges them in.
No stronger teacher. The same model that does the tasks writes the revisions, which separates this from distillation-based adaptation.
Results. On Harvey LAB's Firm Knowledge benchmark, the held-out all-pass rate rises from 30.0% to 55.0% with DeepSeek-V4.1-Flash and from 35.0% to 46.0% with Qwen3.8-Flash.
Scope. Experiments cover one legal domain. Extending self-improvement to the model weights and harness is stated as ongoing work.
Abstract
Foundation models in changing environments can encounter unfamiliar domain knowledge that their original training did not cover. However, new domains rarely provide large amounts of expert-created training data or reliable supervisory signals for further training. We introduce Recursa, a foundation model system that adapts to new domains by learning from its own experience without human experts or a stronger teacher model. Recursa uses a knowledge repository to store what it learns: domain-specific skills and instructions on when to apply them, along with memories of past task experiences and reflections on what to improve. Recursa bootstraps the repository from domain materials and then refines it through recursive self-improvement. It uses parallel learners to execute tasks and synthesize local improvements, then validates and merges the improvements into the knowledge repository. On Harvey LAB's Firm Knowledge benchmark, Recursa raises the all-pass rate on held-out test tasks from 30.0% to 55.0% with DeepSeek-V4.1-Flash and from 35.0% to 46.0% with Qwen3.8-Flash.