[2026-06-07] Issue Resolution — 2 paper(s)

[Zhaoyang-Chu]

📋 2026-06-07 · Issue Resolution — 2 paper(s)

Auto-processed by the arXiv crawler pipeline. Review each paper and reply with the commands below.

---

1. Coherence Collapse: Diagnosing Why Code Agents Fail After Reaching the Right Code

Authors: Myeongsoo Kim, Dingmin Wang, Siwei Cui, Farima Farmahinifarahani, Terry Yue Zhuo, et al.
Venue: arXiv 2026/03 | empirical

The paper introduces TRAJEVAL to decompose agent trajectories into search, read, and edit stages for diagnosing failures in software engineering tasks. It identifies 'Coherence Collapse' as a primary failure mode where agents reach the correct code but subsequently overwrite or destroy it.
📄 Paper

---

2. Rethinking the Value of Agent-Generated Tests for LLM-Based Software Engineering Agents

Authors: Zhi Chen, Zhensu Sun, Yuling Shi, Chao Peng, Xiaodong Gu, et al.
Venue: arXiv 2026/02 | empirical

This paper empirically analyzes the utility of agent-generated tests in repository-level issue resolution using trajectories from six LLMs on SWE-bench Verified. The authors find that while test writing is common, it primarily serves as an observational feedback channel rather than a validation mechanism, and prompt-induced changes in test-writing frequency do not significantly improve performance.
📄 Paper

---

Review commands (comment on this issue):

• /approve all — accept all papers
• /approve 1,3 — accept papers 1 and 3
• /reject 2 — discard paper 2
• /approve 1,3 /reject 2 — mixed
• /edit 1 category=code_generation — change category before approving
• /edit 1 venue=ICSE 2026 — fix venue