Paper page - Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills

Source: https://huggingface.co/papers/2606.07412

Abstract

Socratic-SWE enables self-evolving software engineering agents by leveraging historical solving traces to generate targeted repair tasks that improve agent performance through iterative refinement.

LLM-driven software engineering agentshave become a central testbed for real-world language-model capability, yet their training remains limited by the availability of high-quality SWE tasks. Existingsynthetic data methodstypically create tasks throughfixed mutationorbug-injection procedures, making the resulting distributions largely independent of the agent’s own weaknesses and training progress. We introduce Socratic-SWE, aclosed-loop self-evolution frameworkthat reuses the agent’shistorical solving tracesas a source of training signal. Rather than treating traces only as evidence for reward computation, Socratic-SWE distills them intostructured agent skillsthat summarize recurring failures and effectiverepair patterns. These skills then guide the generation of targeted repair tasks in real repositories. Candidate tasks are checked throughexecution-based validationand scored with asolver-gradient alignment reward, so that the retained tasks are both verifiable and useful for improving the Solver. The updated Solver produces new traces, enabling thetask curriculumto adapt over successive rounds. AcrossSWE-bench Verified,SWE-bench Lite,SWE-bench Pro, andTerminal-Bench 2.0, Socratic-SWE consistently improves overself-evolving baselinesunder the same compute budget, reaching 50.40% onSWE-bench Verifiedafter three iterations. These results suggest that solving traces can serve as a scalable substrate for self-evolving SWE agents.

View arXiv pageView PDFAdd to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.07412 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.07412 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.07412 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.