The paper investigates how verifier feedback influences self‑repair in formal theorem proving. Using controlled one‑step experiments on Lean with frozen Qwen3.6‑27B and DeepSeek‑Prover‑V2‑7B, the authors find that showing a model its failed predecessor can hurt success and increase copying, while providing the corresponding compiler error can partially recover performance. In sequential trials, Qwen benefits from correct diagnostics within failed‑proof histories, but no feedback condition outperforms simple instruction‑only resampling, indicating that verifier feedback can aid repair yet cannot fully offset the cost of conditioning on a failed attempt.
By Jack Li, Sophia Li
The paper introduces a compiler-guided adaptive proof search framework for Lean 4 theorem proving, addressing the challenge of context-dependent proofs in real-world projects. It balances exploration and exploitation by generating diverse starting points via dual-model generation and resampling when stagnation occurs, while refining promising states using compiler-grounded pairwise comparison. Experiments on seven Lean 4 projects from miniCTX‑v2 demonstrate that the method improves average pass rates by 12.8 percentage points within a pass@32 budget and reduces LLM calls by 21.9 % compared to pass@k baselines.
By Zhuo Liu, Ding Yu, Hangfeng He
RePro is a framework that rewrites benchmark problems for large language models (LLMs) in mathematical problem solving, ensuring that the rewritten problems and their answers are valid and correct through Lean-verified proofs. It integrates Lean-oriented neural automated theorem provers (ATPs) to regenerate answers, achieving 100% well-definedness, feasibility, and answer correctness on GSM8K and MATH datasets. Experiments show that models’ performance drops on these proof‑verified rewritten benchmarks, indicating sensitivity to surface‑level and structural variations and potential memorization effects.
By Xiyuan Zhou, Zhuoqi Li, Xinlei Wang, Yirui He, Yuhao Wu, Yuheng Cheng, Yan Xu, Junhua Zhao, Jinjin Gu
arXiv:2606. 29493v1 Announce Type: new Abstract: Benchmarks for LLM-assisted theorem proving in Lean are often treated as intrinsically reliable because every solved instance comes with a machine-checked proof.
By Pawan Sasanka Ammanamanchi, Siddharth Bhat, Stella Biderman
The paper introduces a domain‑specific debugging agent for repairing near‑miss accelerator kernels, arguing that debugging is more efficient than regenerating kernels from scratch. It tackles knowledge scarcity, integrity, and cost control through pattern retrieval, anti‑cheat detection, and convergence safeguards. Empirical results show the debug agent achieves a 66.7% Pass@1 rate and uses 92.8% fewer tokens per success compared to traditional regeneration methods.
By Yansong Sun, Shenxiu Wu, Siyuan Chen, Runlin Hou, Junhao Qiu, Junming Cao, Shudi Shao, Zhichao Lu, Qingfu Zhang
arXiv:2607. 19843v1 Announce Type: cross Abstract: Large language models (LLMs) have made automated program repair (APR) increasingly practical for real-world bugs, but repairing directly from bug reports remains underconstrained.
By Yuhao Tan, Zhibang Yang, Fangkai Yang, Yuan Yao, Yu Kang, Lu Wang, Pu Zhao, Xin Zhang, Xiaoxing Ma, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang
arXiv:2610.17697v1 Announce Type: cross
Abstract: AI systems now write long and difficult mathematical proofs, and checking their step-by-step validity is becoming a bottleneck. A learned verifier co...
By Rodney Lafuente-Mercado, Salman Rahman, Jingyan Shen, James Shiffer, Saadia Gabriel, Pavel Izmailov
arXiv:2609.38384v1 Announce Type: new
Abstract: Verified proof edits offer a natural source of supervision for improving language-model-generated Lean proofs. Yet verification establishes that an edi...
By Pauline Bourigault
arXiv:2608. 02712v1 Announce Type: cross Abstract: Kernel generation for hardware accelerators such as GPUs and NPUs has become a proving ground for large language models (LLMs), and state-of-the-art systems raise correctness through pipelines that couple LLMs with agentic reinforcement learning and evolutionary search.
By Yansong Sun, Shenxiu Wu, Siyuan Chen, Runlin Hou, Junhao Qiu, Junming Cao, Shudi Shao, Zhichao Lu, Qingfu Zhang
arXiv:2607. 22972v1 Announce Type: new Abstract: Lean~4's \grind{} tactic combines congruence closure, \ematch{}ing, and case-splitting into a single automated solver, and like any such solver, it relies on hand-tuned heuristics to decide what to instantiate and where to case-split.
By Evan Wang, Simon Chess, Sophie Szeto, Theodore Meek
arXiv:2607. 12605v1 Announce Type: cross Abstract: Large language models (LLMs) have improved automated program repair (APR), but two limitations remain.
By Zhili Huang, Ling Xu, Hongyu Zhang
arXiv:2607. 17641v1 Announce Type: new Abstract: Verify-repair loops are a standard means for large language model (LLM) agents to correct faulty plans in code generation, mathematical reasoning, and tool use.
By Yitao Wu, Si Shen, Rui Yang, Hong Peng, Bin Hu