“My stack lately has been Next.js with Prettier and very strict lint, type, and complexity gates, plus opinionated tests that aim for real signal over mere coverage theatre. I want to find the time to fine-tune GPT-4o using before/after examples of code that fails the gates and then passes them. The hope being it generates gate-compliant code on the first…”
RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning
Large language models (LLMs) deployed as agents solve user-specified tasks over multiple steps while keeping the required manual engagement to a minimum. Crucially, such LLMs need to ground their generations in any feedback obtained to reliably achieve the desired outcomes. We propose an end-to-end reinforcement learning method for teaching models to leverage execution feedback in the realm of code synthesis, where state-of-the-art LLMs struggle to improve code iteratively compared to independent sampling. We benchmark on competitive programming tasks, where we achieve new state-of-the art results with both small (8B parameters) and large (70B) models while reducing the amount of samples required by an order of magnitude. Our analysis of inference-time behavior demonstrates that our method produces LLMs that effectively leverage automatic feedback over multiple steps.
候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及- RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning ↗jasondavies · 2024-10-04 · 4 分