SWE-Ladder：通过渐进式特权自蒸馏学习定位、复现与修复

View PDF HTML (experimental)

Abstract:Resolving a repository-level issue typically requires a software-engineering (SWE) agent to localize the faulty code, reproduce the failure, and patch the code. Reinforcement learning trains such agents from a single test result on the final patch, while on-policy self-distillation (OPSD) adds token-level supervision from a teacher with privileged evidence, exposing the full reference patch at every step. Neither provides stage-specific guidance for localization or reproduction. We propose SWE-Ladder, which turns the verified repair of each training instance into stage-specific evidence: the changed files and lines for localization, the fail-to-pass tests for reproduction, and the diff for patching. A rule-based router assigns each step of an end-to-end rollout to a stage based on its tool call, a frozen teacher sees only the matching evidence, and the stages are distilled in repair order while GRPO optimizes the full trajectory. With Qwen3-4B-Instruct-2507 on SWE-bench Verified under the R2E-Gym scaffold, SWE-Ladder improves pass@1 from 9.6% to 16.4%, compared with 10.4% for GRPO and 12.4% for OPSD with the full patch. Under the OpenHands scaffold, it improves pass@1 from 13.0% to 17.6%. Reversing the stage order under R2E-Gym reduces pass@1 to 11.4%.

Comments:	18 pages, 11 figures. Code: this https URL
Subjects:	Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
Cite as:	arXiv:2610.33667 [cs.AI]
	(or arXiv:2610.33667v1 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2610.33667 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Xudong Cai [view email]
[v1] Sat, 24 Oct 2026 13:27:18 UTC (3,767 KB)