Metraipolis › archive › dc_bbab233f3704
Improving Large Language Models for Code through Runtime Program-State Reasoning — arXiv:2609.34359
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34359
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:57:42.780Z
- SHA-256 of the archived text
0beee109f5f76cc8aa7faca486fefee063a4ccc46f21468cf88f00114a448c3e- Stable URL of this page
https://metraipolis.com/documento/dc_bbab233f3704
How to cite
Improving Large Language Models for Code through Runtime Program-State Reasoning — arXiv:2609.34359 — arXiv (en), https://arxiv.org/abs/2609.34359 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bbab233f3704 (Metraipolis archive dc_bbab233f3704, sha256 0beee109f5f76cc8aa7faca486fefee063a4ccc46f21468cf88f00114a448c3e)
BibTeX
@misc{dcbbab233f3704,
title = {Improving Large Language Models for Code through Runtime Program-State Reasoning — arXiv:2609.34359},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34359},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bbab233f3704; sha256 0beee109f5f76cc8aa7faca486fefee063a4ccc46f21468cf88f00114a448c3e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34359. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models receive limited explicit training in reasoning about runtime program states. We study whether training models to reason about runtime program states improves downstream software-engineering capabilities. We introduce two complementary program-state reasoning tasks. Buggy input-output reasoning requires a model to generate a concrete input that exposes a behavioral difference between a buggy program and a hidden correct implementation and to predict the resulting execution behavior. Precondition-postcondition reasoning requires an agent to symbolically characterize a bug-triggering precondition, predict the expected postcondition, explain their causal connection, and instantiate this reasoning as an executable regression test. By incorporating these two tasks into a staged post-training pipeline, we develop Comet-9B, a 9B language model based on Qwen3.5-9B Base. We evaluate the resulting checkpoints on repository-level patch generation, regression-test generation, and security PoC generation. Adding both program-state reasoning tasks to supervised fine-tuning (SFT) on issue resolution improves success rates by 7.25 percentage points on SWE-bench Pro and 9.70 points on SWT-Bench Verified. Sequential reinforcement learning on the two tasks yields further gains of 7.25, 26.79, and 4.67 percentage points on SWE-bench Pro, SWT-Bench Verified, and CyberGym, respectively. Despite having only 9B parameters, Comet-9B achieves a score comparable to the reported GPT-5.2 result on SWE-bench Pro and matches the reported success rate of a GPT-4o-based agent on SWT-Bench Verified. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.34359 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T20:29:01.530Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).