Tag
斯坦福研究者通过受控实验研究了在Lean定理证明中，验证器反馈（失败证明与编译诊断）对模型自我修复的影响，发现虽然诊断能改善修复，但自我修复整体上仍无法超越独立重采样。
A journalist successfully uses John Deere's self-repair software to fix a tractor, but farmers remain unconvinced due to the company's history of restricting third-party repairs.
Fairphone's Gen 6+ smartphone is designed for ethics, repairability, and sustainability, featuring easy disassembly and long-term support to challenge conventional smartphone manufacturing.
This paper introduces MERIT, a training-free agent that uses causal episodic memory of past repair outcomes to improve subsequent Text-to-SQL generations, boosting execution accuracy on Spider and BIRD benchmarks.
SAAG proposes a cascaded diagnostic framework for evaluating LLM agent function calling by decomposing evaluation into registry conformance, structural completeness, and argument grounding stages, enabling interpretable diagnostics and iterative self-repair. Experiments with sub-4B models show improved argument precision and reduced value hallucination compared to single-pass evaluation.
A new agent specification framework built in Rust, featuring a small binary size, Vuepress integration for documentation, issue persistence for improved code review context, and self-repair capabilities. Compatible with several agent tools and still under development.