执行反馈可以指导编码代理进行正确的存储库修复,，但前提是测试捕获了问题所请求的行为。当相同的轨迹写入补丁和测试,时，代理生成的测试可以编码不完整或不正确的行为目标;，它们的错误可能会一致并产生错误的置信度。我们引入了 ExecCritic,，将测试-验证-修订支架与用于在其中训练代理的特定于角色的强化学习配方相结合。脚手架将测试构建与源代码修复分开:，测试代理独立生成存储库本机测试,，故障关闭线束合格并冻结它们,，修复代理根据执行反馈修改源代码，而不更改测试。两个角色都使用 Qwen-3.5-35B-A3B 作为骨干，并分别进行训练。在学习测试, 中，测试代理学习生成行为有效的测试，以区分正确的补丁和不正确的补丁。在“改进, 测试”中，修复代理学习直接任务解决和反馈引导的修订。在 SWE 平台上，经过验证的, 测试质量决定了反馈是否有助于: 保持基本修复代理固定, 基本测试代理测试将解决率从无测试基线 61.2% 降低到 57.3%,，而 GPT-5.6-sol 的测试将其提高到 65.3%。特定于角色的训练后将 Qwen 测试代理的 的基础到黄金成功率从 22.2% 提高到 62.2%;，两个训练后的 Qwen 代理达到 72.6%,，比原始无测试基线提高了 11.4 点，在评估时没有更强的模型或 Oracle 反馈。代码可在此 https URL 公开获取。

Execution feedback can guide coding agents toward correct repository repairs, but only when the tests capture the behavior requested by the issue. Agent-generated tests can encode incomplete or incorrect behavioral targets; when the same trajectory writes both the patch and the test, their errors can agree and create false confidence. We introduce ExecCritic, combining a test--verify--revise scaffold with a role-specific reinforcement learning recipe for training agents within it. The scaffold separates test construction from source-code repair: a Test agent independently generates repository-native tests, a fail-closed harness qualifies and freezes them, and a Repair agent revises source code from their execution feedback without changing the tests. Both roles use Qwen-3.5-35B-A3B as the backbone and are trained separately. In Learn to Test, the Test agent learns to produce behaviorally valid tests that distinguish correct from incorrect patches. In Test to Improve, the Repair agent learns both direct task resolution and feedback-guided revision. On SWE-bench Verified, test quality determines whether feedback helps: holding the base Repair agent fixed, tests from the base Test agent reduce resolved rate from a no-test baseline of 61.2% to 57.3%, whereas tests from GPT-5.6-sol raise it to 65.3%. Role-specific post-training raises the Qwen Test agent的 Base-to-Gold success from 22.2% to 62.2%; composing the two post-trained Qwen agents reaches 72.6%, an 11.4-point gain over the original no-test baseline without stronger-model or Oracle feedback at evaluation time. Code is publicly available at this https URL.

科目: 人工智能 (cs.AI); 计算和语言 (cs.CL); 软件工程 (cs.SE)

Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)