Teaching Your Models to Understand Code via Focal Preference Alignment

• URL: http://arxiv.org/abs/2503.02783v4
• Date: Thu, 09 Oct 2025 07:51:19 GMT
• Title: Teaching Your Models to Understand Code via Focal Preference Alignment
• Authors: Jie Wu, Haoling Li, Xin Zhang, Xiao Liu, Yangyu Huang, Jianwen Luo, Yizhen Zhang, Zuchao Li, Ruihang Chu, Yujiu Yang, Scarlett Li,
• Abstract summary: In existing approaches, a set of n candidate solutions is evaluated based on test case success rates.<n>Because this approach aligns entire failing code blocks rather than pinpointing specific errors, it lacks the granularity necessary to capture meaningful error-correction relationships.<n>We propose Target-DPO, a new preference alignment framework that mimics human iterative debug to refine Code LLMs.
• Score: 70.71693365502212
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Preference learning extends the performance of Code LLMs beyond traditional supervised fine-tuning by leveraging relative quality comparisons. In existing approaches, a set of n candidate solutions is evaluated based on test case success rates, with the candidate demonstrating a higher pass rate being labeled as positive and its counterpart with a lower pass rate as negative. However, because this approach aligns entire failing code blocks rather than pinpointing specific errors, it lacks the granularity necessary to capture meaningful error-correction relationships. As a result, the model is unable to learn more informative error-correction patterns. To address these issues, we propose Target-DPO, a new preference alignment framework that mimics human iterative debugging to refine Code LLMs. Target-DPO explicitly locates error regions and aligns the corresponding tokens via a tailored DPO algorithm. To facilitate it, we introduce the CodeFlow dataset, where samples are iteratively refined until passing tests, with modifications capturing error corrections. Extensive experiments show that a diverse suite of Code LLMs equipped with Target-DPO achieves significant performance gains in code generation and improves on challenging tasks like BigCodeBench. In-depth analysis reveals that Target-DPO yields fewer errors. Code, model and datasets are in: https://github.com/JieWu02/Target-DPO.

Related papers

• Towards Automated Error Discovery: A Study in Conversational AI [48.735443116662026]
  We introduce Automated Error Discovery, a framework for detecting and defining errors in conversational AI.<n>We also propose SEEED (Soft Clustering Extended-Based Error Detection), as an encoder-based approach to its implementation.
  arXiv  Detail & Related papers  (2025-09-13T14:53:22Z)
• Alignment with Fill-In-the-Middle for Enhancing Code Generation [56.791415642365415]
  We propose a novel approach that splits code snippets into smaller, granular blocks, creating more diverse DPO pairs from the same test cases.<n>Our approach demonstrates significant improvements in code generation tasks, as validated by experiments on benchmark datasets such as HumanEval (+), MBPP (+), APPS, LiveCodeBench, and BigCodeBench.
  arXiv  Detail & Related papers  (2025-08-27T03:15:53Z)
• Understanding and Mitigating Errors of LLM-Generated RTL Code [7.747889860813149]
  Large language model (LLM) based register-transfer-level (RTL) code generation is promising but the overall success rate remains unsatisfactory.<n>We conduct a comprehensive error analysis and manual categorization.<n>Our findings reveal that most errors stem from insufficient RTL programming knowledge, poor understanding of circuit concepts, or misinterpretation of complex multimodal inputs.
  arXiv  Detail & Related papers  (2025-08-07T11:02:32Z)
• CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward [50.97588334916863]
  We develop CompassVerifier, an accurate and robust lightweight verifier model for evaluation and outcome reward.<n>It demonstrates multi-domain competency spanning math, knowledge, and diverse reasoning tasks, with the capability to process various answer types.<n>We introduce VerifierBench benchmark comprising model outputs collected from multiple data sources, augmented through manual analysis of metaerror patterns to enhance CompassVerifier.
  arXiv  Detail & Related papers  (2025-08-05T17:55:24Z)
• KodCode: A Diverse, Challenging, and Verifiable Synthetic Dataset for Coding [49.56049319037421]
  KodCode is a synthetic dataset that addresses the persistent challenge of acquiring high-quality, verifiable training data. It comprises question-solution-test triplets that are systematically validated via a self-verification procedure. This pipeline yields a large-scale, robust and diverse coding dataset.
  arXiv  Detail & Related papers  (2025-03-04T19:17:36Z)
• Focused-DPO: Enhancing Code Generation Through Focused Preference Optimization on Error-Prone Points [51.40935517552926]
  We introduce Focused-DPO, a framework that enhances code generation by directing preference optimization towards critical error-prone areas.<n>By focusing on error-prone points, Focused-DPO advances the accuracy and functionality of model-generated code.
  arXiv  Detail & Related papers  (2025-02-17T06:16:02Z)
• Are LLMs Better than Reported? Detecting Label Errors and Mitigating Their Effect on Model Performance [28.524573212179124]
  Large language models (LLMs) offer new opportunities to enhance the annotation process.<n>We compare expert, crowd-sourced, and LLM-based annotations in terms of the agreement, label quality, and efficiency.<n>Our findings reveal a substantial number of label errors, which, when corrected, a significant upward shift in reported model performance.
  arXiv  Detail & Related papers  (2024-10-24T16:27:03Z)
• Subtle Errors Matter: Preference Learning via Error-injected Self-editing [59.405145971637204]
  We propose a novel preference learning framework called eRror-Injected Self-Editing (RISE)<n>RISE injects predefined subtle errors into pivotal tokens in reasoning or steps to construct hard pairs for error mitigation.<n>Experiments validate the effectiveness of RISE, with preference learning on Qwen2-7B-Instruct yielding notable improvements of 3.0% on GSM8K and 7.9% on MATH with only 4.5K training samples.
  arXiv  Detail & Related papers  (2024-10-09T07:43:38Z)
• Insights from Benchmarking Frontier Language Models on Web App Code Generation [1.7268889851975326]
  This paper presents insights from evaluating 16 frontier large language models (LLMs) on the WebApp1K benchmark. The results reveal that while all models possess similar underlying knowledge, their performance is differentiated by the frequency of mistakes they make.
  arXiv  Detail & Related papers  (2024-09-08T18:24:26Z)
• COAST: Enhancing the Code Debugging Ability of LLMs through Communicative Agent Based Data Synthesis [29.667170755786508]
  We introduce EVAL, a benchmark for evaluating the abilities of Large Language Models.<n>We propose the COmmunicative Agent-based data SynThesis framework, which employs a multi-agent system to generate high-quality training data.<n>Results demonstrate that COAST-generated data outperform human-curated and GPT-4-generated data.
  arXiv  Detail & Related papers  (2024-08-09T11:35:44Z)
• GLARE: Low Light Image Enhancement via Generative Latent Feature based Codebook Retrieval [80.96706764868898]
  We present a new Low-light Image Enhancement (LLIE) network via Generative LAtent feature based codebook REtrieval (GLARE) We develop a generative Invertible Latent Normalizing Flow (I-LNF) module to align the LL feature distribution to NL latent representations, guaranteeing the correct code retrieval in the codebook. Experiments confirm the superior performance of GLARE on various benchmark datasets and real-world data.
  arXiv  Detail & Related papers  (2024-07-17T09:40:15Z)
• CodecLM: Aligning Language Models with Tailored Synthetic Data [51.59223474427153]
  We introduce CodecLM, a framework for adaptively generating high-quality synthetic data for instruction-following abilities. We first encode seed instructions into metadata, which are concise keywords generated on-the-fly to capture the target instruction distribution. We also introduce Self-Rubrics and Contrastive Filtering during decoding to tailor data-efficient samples.
  arXiv  Detail & Related papers  (2024-04-08T21:15:36Z)
• SEED: Customize Large Language Models with Sample-Efficient Adaptation for Code Generation [35.88318116340547]
  We propose a novel adaptation approach named SEED, which stands for Sample-Efficient adaptation with Error-Driven learning for code generation. We show that SEED achieves superior performance with few training samples, showing an average relative improvement of 54.7% in Pass@1 on multiple code generation benchmarks.
  arXiv  Detail & Related papers  (2024-02-29T16:09:02Z)
• Learning to Check: Unleashing Potentials for Self-Correction in Large Language Models [5.463333911506443]
  We aim to enhance the self-checking capabilities of large language models (LLMs) by constructing training data for checking tasks. We propose a specialized checking format called "Step CoT Check" Experiments demonstrate that fine-tuning with the "Step CoT Check" format significantly improves the self-checking and self-correction abilities of LLMs.
  arXiv  Detail & Related papers  (2024-02-20T14:23:23Z)
• StepCoder: Improve Code Generation with Reinforcement Learning from Compiler Feedback [58.20547418182074]
  We introduce StepCoder, a novel framework for code generation, consisting of two main components. CCCS addresses the exploration challenge by breaking the long sequences code generation task into a Curriculum of Code Completion Subtasks. FGO only optimize the model by masking the unexecuted code segments to provide Fine-Grained Optimization. Our method improves the ability to explore the output space and outperforms state-of-the-art approaches in corresponding benchmarks.
  arXiv  Detail & Related papers  (2024-02-02T13:14:31Z)
• Understanding and Mitigating Classification Errors Through Interpretable Token Patterns [58.91023283103762]
  Characterizing errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors. We propose to discover those patterns of tokens that distinguish correct and erroneous predictions. We show that our method, Premise, performs well in practice.
  arXiv  Detail & Related papers  (2023-11-18T00:24:26Z)
• Fixing Large Language Models' Specification Misunderstanding for Better Code Generation [13.494822086550604]
  muFiX is a novel prompting technique to improve the code generation performance of large language models (LLMs) It first exploits test case analysis to obtain specification understanding and enables a self-improvement process. muFiX further fixes the specification understanding towards the direction reducing the gap between the provided understanding and the actual understanding.
  arXiv  Detail & Related papers  (2023-09-28T02:58:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.