Aligning CodeLLMs with Direct Preference Optimization
원문 초록
The last year has witnessed the rapid progress of large language models (LLMs) across diverse domains. Among them, CodeLLMs have garnered particular attention because they can not only assist in completing various programming tasks but also represent the decision-making and logical reasoning capabilities of LLMs. However, current CodeLLMs mainly focus on pre-training and supervised fine-tuning scenarios, leaving the alignment stage, which is important for post-training LLMs, under-explored. This work first identifies that the commonly used PPO algorithm may be suboptimal for the alignment of CodeLLM because the involved reward rules are routinely coarse-grained and potentially flawed. We then advocate addressing this using the DPO algorithm. Based on only preference data pairs, DPO can render the model rank data automatically, giving rise to a fine-grained rewarding pattern more robust than human intervention. We also contribute a pipeline for collecting preference pairs for DPO on CodeLLMs. Studies show that our method significantly improves the performance of existing CodeLLMs on benchmarks such as MBPP and HumanEval.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
any4: Learned 4-bit Numeric Representation for LLMs
Self-Correcting Code Generation Using Small Language Models
SACL: Understanding and Combating Textual Bias in Code Retrieval with Semantic-Augmented Reranking and Localization
Plan for Speed -- Dilated Scheduling for Masked Diffusion Language Models
CAPO: Reinforcing Consistent Reasoning in Medical Decision-Making
Guideline Forest: Experience-Induced Multi-Guideline Reasoning with Stepwise Aggregation
토론
이 논문과 연결된 포럼 글이 아직 없습니다.