titleTermination Lag: Coding Agents Solve Early but Stop Late
authorsMusa Cim, Burak Topcu, Mahmut Taylan Kandemir
date2026-11-11
venuearXiv (preprint, not peer reviewed)
idarXiv:2611.14960
affiliationPennsylvania State University
upvotes7 (Hugging Face Daily Papers, 2026-11-19)
linksabs pdf hf code
cite
relatedProgress-Aware Distillation for Mitigating Stagnation in Small Language Model Agents
SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops
ATLAS: Aligned Transport of Latent Structure for Reliable World Model Planning
Agent Against Agent: An Agentic System for Automatic Prompt Injection Red Teaming
FlashRT: Towards Computationally and Memory Efficient Red-Teaming for Prompt Injection and Knowledge Corruption
abstractCoding agents can solve a task well before they stop working on it. We characterize termination lag--the number of steps between an agent's first correct patch and the end of its run--across five models, two agent frameworks, and three SWE-bench datasets by recording and testing every distinct repository patch generated during execution. Our analysis reveals that 38--82% of successful runs achieve a correct solution on their initial source-code edit. However, agents continue to expend substantial effort after correct patches, with post-solution steps accounting for a median of 35--37% for Qwen3-Coder and 54--60% for Laguna-XS.2 in successful runs on SWE-bench Verified and Multilingual. Agents spend most of these additional steps verifying their work and exploring the repository. In a small fraction of cases (1--8%), this continued activity leads agents to undo their earlier correct patch, leaving the final solution incorrect.

bibtex

@misc{cim2026termination,
  title={Termination Lag: Coding Agents Solve Early but Stop Late},
  author={Musa Cim and Burak Topcu and Mahmut Taylan Kandemir},
  year={2026},
  eprint={2611.14960},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2611.14960}
}