In this report, we introduce the Qwen2.5-Coder series, a significant upgrade from its predecessor, CodeQwen1.5. This series includes six models: Qwen2.5-Coder-(0.5B/1.5B/3B/7B/14B/32B). As a code-specific model, Qwen2.5-Coder is built upon the Qwen2.5 architecture and continues pretrained on a vast corpus of over 5.5 trillion tokens. Through meticulous data cleaning, scalable synthetic data generation, and balanced data mixing, Qwen2.5-Coder demonstrates impressive code generation capabilities while retaining general and math skills. These models have been evaluated on a wide range of code-related tasks, achieving state-of-the-art (SOTA) performance across more than 10 benchmarks, including code generation, completion, reasoning, and repair, consistently outperforming larger models of the same model size. We believe that the release of the Qwen2.5-Coder series will advance research in code intelligence and, with its permissive licensing, support wider adoption by developers in real-world applications.
Recent advancements in automatic code generation using large language models (LLMs) have brought us closer to fully automated secure software development. However, existing approaches often rely on a single agent for code generation, which struggles to produce secure, vulnerability-free code. Traditional program synthesis with LLMs has primarily focused on functional correctness, often neglecting critical dynamic security implications that happen during runtime. To address these challenges, we propose AutoSafeCoder, a multi-agent framework that leverages LLM-driven agents for code generation, vulnerability analysis, and security enhancement through continuous collaboration. The framework consists of three agents: a Coding Agent responsible for code generation, a Static Analyzer Agent identifying vulnerabilities, and a Fuzzing Agent performing dynamic testing using a mutation-based fuzzing approach to detect runtime errors. Our contribution focuses on ensuring the safety of multi-agent code generation by integrating dynamic and static testing in an iterative process during code generation by LLM that improves security. Experiments using the SecurityEval dataset demonstrate a 13% reduction in code vulnerabilities compared to baseline LLMs, with no compromise in functionality.
Selecting the best code solution from multiple generated ones is an essential task in code generation, which can be achieved by using some reliable validators (e.g., developer-written test cases) for assistance. Since reliable test cases are not always available and can be expensive to build in practice, researchers propose to automatically generate test cases to assess code solutions. However, when both code solutions and test cases are plausible and not reliable, selecting the best solution becomes challenging. Although some heuristic strategies have been proposed to tackle this problem, they lack a strong theoretical guarantee and it is still an open question whether an optimal selection strategy exists. Our work contributes in two ways. First, we show that within a Bayesian framework, the optimal selection strategy can be defined based on the posterior probability of the observed passing states between solutions and tests. The problem of identifying the best solution is then framed as an integer programming problem. Second, we propose an efficient approach for approximating this optimal (yet uncomputable) strategy, where the approximation error is bounded by the correctness of prior knowledge. We then incorporate effective prior knowledge to tailor code generation tasks. Both theoretical and empirical studies confirm that existing heuristics are limited in selecting the best solutions with plausible test cases. Our proposed approximated optimal strategy B4 significantly surpasses existing heuristics in selecting code solutions generated by large language models (LLMs) with LLM-generated tests, achieving a relative performance improvement by up to 50% over the strongest heuristic and 246% over the random selection in the most challenging scenarios. Our code is publicly available at https://github.com/ZJU-CTAG/B4.
Reinforcement learning from human feedback (RLHF) is one of the key techniques that helps large language models (LLMs) to follow instructions and provide helpful and harmless responses. While direct policy optimization methods exist, state-of-the-art LLMs adopt RL-based methods (usually PPO) in RLHF to train the policy to generate good responses guided by a reward model learned from preference data. The main challenge of these methods is the inaccuracy of the intermediate reward model, especially in code generation tasks that require long and complex reasoning to score a response. We find that the reliability of the reward model varies across responses assigned with different rewards. This motivates us to filter the samples whose rewards may be unreliable to improve signal-to-noise ratio during policy learning, resulting in Policy Filtration for Proximal Policy Optimization (PF-PPO). To choose a proper policy filtration strategy for a given reward model, the coefficient of determination ($R^2$) between rewards and actual scores on filtered samples serves as a good metrics and helps us find several promising strategies. We provide extensive experiments to validate the effectiveness of PF-PPO in code generation tasks, and find that some variants of PF-PPO are highly effective and achieve new state-of-the-art performance across 7-billion-parameter models on HumanEval, MBPP, and a new and more challenging LeetCode Contest benchmark.
Large language models (LLMs) have shown great potential for automatic code generation and form the basis for various tools such as GitHub Copilot. However, recent studies highlight that many LLM-generated code contains serious security vulnerabilities. While previous work tries to address this by training models that generate secure code, these attempts remain constrained by limited access to training data and labor-intensive data preparation. In this paper, we introduce HexaCoder, a novel approach to enhance the ability of LLMs to generate secure codes by automatically synthesizing secure codes, which reduces the effort of finding suitable training data. HexaCoder comprises two key components: an oracle-guided data synthesis pipeline and a two-step process for secure code generation. The data synthesis pipeline generates pairs of vulnerable and fixed codes for specific Common Weakness Enumeration (CWE) types by utilizing a state-of-the-art LLM for repairing vulnerable code. A security oracle identifies vulnerabilities, and a state-of-the-art LLM repairs them by extending and/or editing the codes, creating data pairs for fine-tuning using the Low-Rank Adaptation (LoRA) method. Each example of our fine-tuning dataset includes the necessary security-related libraries and code that form the basis of our novel two-step generation approach. This allows the model to integrate security-relevant libraries before generating the main code, significantly reducing the number of generated vulnerable codes by up to 85% compared to the baseline methods. We perform extensive evaluations on three different benchmarks for four LLMs, demonstrating that HexaCoder not only improves the security of the generated code but also maintains a high level of functional correctness.
This paper presents insights from evaluating 16 frontier large language models (LLMs) on the WebApp1K benchmark, a test suite designed to assess the ability of LLMs to generate web application code. The results reveal that while all models possess similar underlying knowledge, their performance is differentiated by the frequency of mistakes they make. By analyzing lines of code (LOC) and failure distributions, we find that writing correct code is more complex than generating incorrect code. Furthermore, prompt engineering shows limited efficacy in reducing errors beyond specific cases. These findings suggest that further advancements in coding LLM should emphasize on model reliability and mistake minimization.
Large language models (LLMs) have achieved impressive performance on code generation. Although prior studies enhanced LLMs with prompting techniques and code refinement, they still struggle with complex programming problems due to rigid solution plans. In this paper, we draw on pair programming practices to propose PairCoder, a novel LLM-based framework for code generation. PairCoder incorporates two collaborative LLM agents, namely a Navigator agent for high-level planning and a Driver agent for specific implementation. The Navigator is responsible for proposing promising solution plans, selecting the current optimal plan, and directing the next iteration round based on execution feedback. The Driver follows the guidance of Navigator to undertake initial code generation, code testing, and refinement. This interleaved and iterative workflow involves multi-plan exploration and feedback-based refinement, which mimics the collaboration of pair programmers. We evaluate PairCoder with both open-source and closed-source LLMs on various code generation benchmarks. Extensive experimental results demonstrate the superior accuracy of PairCoder, achieving relative pass@1 improvements of 12.00%-162.43% compared to prompting LLMs directly.
Recently, there has been a growing interest in studying how to construct better code instruction tuning data. However, we observe Code models trained with these datasets exhibit high performance on HumanEval but perform worse on other benchmarks such as LiveCodeBench. Upon further investigation, we find that many datasets suffer from severe data leakage. After cleaning up most of the leaked data, some well-known high-quality datasets perform poorly. This discovery reveals a new challenge: identifying which dataset genuinely qualify as high-quality code instruction data. To address this, we propose an efficient code data pruning strategy for selecting good samples. Our approach is based on three dimensions: instruction complexity, response quality, and instruction diversity. Based on our selected data, we present XCoder, a family of models finetuned from LLaMA3. Our experiments show XCoder achieves new state-of-the-art performance using fewer training data, which verify the effectiveness of our data strategy. Moreover, we perform a comprehensive analysis on the data composition and find existing code datasets have different characteristics according to their construction methods, which provide new insights for future code LLMs. Our models and dataset are released in https://github.com/banksy23/XCoder
Large language models (LLMs) have significantly improved code generation, particularly in one-pass code generation. However, most existing approaches focus solely on generating code in a single programming language, overlooking the potential of leveraging the multi-language capabilities of LLMs. LLMs have varying patterns of errors across different languages, suggesting that a more robust approach could be developed by leveraging these multi-language outputs. In this study, we propose Multi-Programming Language Ensemble (MPLE), a novel ensemble-based method that utilizes code generation across multiple programming languages to enhance overall performance. By treating each language-specific code generation process as an individual "weak expert" and effectively integrating their outputs, our method mitigates language-specific errors and biases. This multi-language ensemble strategy leverages the complementary strengths of different programming languages, enabling the model to produce more accurate and robust code. Our approach can be seamlessly integrated with commonly used techniques such as the reflection algorithm and Monte Carlo tree search to improve code generation quality further. Experimental results show that our framework consistently enhances baseline performance by up to 17.92% on existing benchmarks (HumanEval and HumanEval-plus), with a standout result of 96.25% accuracy on the HumanEval benchmark, achieving new state-of-the-art results across various LLM models. The code will be released at https://github.com/NinjaTech-AI/MPLE
While scaling training compute has led to remarkable improvements in large language models (LLMs), scaling inference compute has not yet yielded analogous gains. We hypothesize that a core missing component is a lack of diverse LLM outputs, leading to inefficient search due to models repeatedly sampling highly similar, yet incorrect generations. We empirically demonstrate that this lack of diversity can be mitigated by searching over candidate plans for solving a problem in natural language. Based on this insight, we propose PlanSearch, a novel search algorithm which shows strong results across HumanEval+, MBPP+, and LiveCodeBench (a contamination-free benchmark for competitive coding). PlanSearch generates a diverse set of observations about the problem and then uses these observations to construct plans for solving the problem. By searching over plans in natural language rather than directly over code solutions, PlanSearch explores a significantly more diverse range of potential solutions compared to baseline search methods. Using PlanSearch on top of Claude 3.5 Sonnet achieves a state-of-the-art pass@200 of 77.0% on LiveCodeBench, outperforming both the best score achieved without search (pass@1 = 41.4%) and using standard repeated sampling (pass@200 = 60.6%). Finally, we show that, across all models, search algorithms, and benchmarks analyzed, we can accurately predict performance gains due to search as a direct function of the diversity over generated ideas. Code can be found at https://github.com/scaleapi/plansearch.
The launch of ChatGPT in November 2022 marked the beginning of a new era in AI, the availability of generative AI tools for everyone to use. ChatGPT and other similar chatbots boast a wide range of capabilities from answering student homework questions to creating music and art. Given the large amounts of human data chatbots are built on, it is inevitable that they will inherit human errors and biases. These biases have the potential to inflict significant harm or increase inequity on different subpopulations. Because chatbots do not have an inherent understanding of societal values, they may create new content that is contrary to established norms. Examples of concerning generated content includes child pornography, inaccurate facts, and discriminatory posts. In this position paper, we argue that the speed of advancement of this technology requires us, as computer and data scientists, to mobilize and develop a values-based auditing framework containing a community established standard set of measurements to monitor the health of different chatbots and LLMs. To support our argument, we use a simple audit template to share the results of basic audits we conduct that are focused on measuring potential bias in search engine style tasks, code generation, and story generation. We identify responses from GPT 3.5 and GPT 4 that are both consistent and not consistent with values derived from existing law. While the findings come as no surprise, they do underscore the urgency of developing a robust auditing framework for openly sharing results in a consistent way so that mitigation strategies can be developed by the academic community, government agencies, and companies when our values are not being adhered to. We conclude this paper with recommendations for value-based strategies for improving the technologies.
Recent advancements in Large Language Models (LLMs) are transforming biology, computer science, engineering, and every day life. However, integrating the wide array of computational tools, databases, and scientific literature continues to pose a challenge to biological research. LLMs are well-suited for unstructured integration, efficient information retrieval, and automating standard workflows and actions from these diverse resources. To harness these capabilities in bioinformatics, we present a prototype Bioinformatics Retrieval Augmented Digital assistant (BRAD). BRAD is a chatbot and agentic system that integrates a variety of bioinformatics tools. The Python package implements an AI \texttt{Agent} that is powered by LLMs and connects to a local file system, online databases, and a user's software. The \texttt{Agent} is highly configurable, enabling tasks such as Retrieval-Augmented Generation, searches across bioinformatics databases, and the execution of software pipelines. BRAD's coordinated integration of bioinformatics tools delivers a context-aware and semi-autonomous system that extends beyond the capabilities of conventional LLM-based chatbots. A graphical user interface (GUI) provides an intuitive interface to the system.
Large language models (LLMs), such as ChatGPT and Copilot, are transforming software development by automating code generation and, arguably, enable rapid prototyping, support education, and boost productivity. Therefore, correctness and quality of the generated code should be on par with manually written code. To assess the current state of LLMs in generating correct code of high quality, we conducted controlled experiments with ChatGPT and Copilot: we let the LLMs generate simple algorithms in Java and Python along with the corresponding unit tests and assessed the correctness and the quality (coverage) of the generated (test) codes. We observed significant differences between the LLMs, between the languages, between algorithm and test codes, and over time. The present paper reports these results together with the experimental methods allowing repeated and comparable assessments for more algorithms, languages, and LLMs over time.
Data analysts often need to iterate between data transformations and chart designs to create rich visualizations for exploratory data analysis. Although many AI-powered systems have been introduced to reduce the effort of visualization authoring, existing systems are not well suited for iterative authoring. They typically require analysts to provide, in a single turn, a text-only prompt that fully describe a complex visualization. We introduce Data Formulator 2 (DF2 for short), an AI-powered visualization system designed to overcome this limitation. DF2 blends graphical user interfaces and natural language inputs to enable users to convey their intent more effectively, while delegating data transformation to AI. Furthermore, to support efficient iteration, DF2 lets users navigate their iteration history and reuse previous designs, eliminating the need to start from scratch each time. A user study with eight participants demonstrated that DF2 allowed participants to develop their own iteration styles to complete challenging data exploration sessions.
Language models (LMs) have exhibited impressive abilities in generating codes from natural language requirements. In this work, we highlight the diversity of code generated by LMs as a critical criterion for evaluating their code generation capabilities, in addition to functional correctness. Despite its practical implications, there is a lack of studies focused on assessing the diversity of generated code, which overlooks its importance in the development of code LMs. We propose a systematic approach to evaluate the diversity of generated code, utilizing various metrics for inter-code similarity as well as functional correctness. Specifically, we introduce a pairwise code similarity measure that leverages large LMs' capabilities in code understanding and reasoning, demonstrating the highest correlation with human judgment. We extensively investigate the impact of various factors on the quality of generated code, including model sizes, temperatures, training approaches, prompting strategies, and the difficulty of input problems. Our consistent observation of a positive correlation between the test pass score and the inter-code similarity score indicates that current LMs tend to produce functionally correct code with limited diversity.
Recently, a diverse set of decoding and reranking procedures have been shown effective for LLM-based code generation. However, a comprehensive framework that links and experimentally compares these methods is missing. We address this by proposing Decoding Objectives for Code Execution, a comprehensive framework that includes candidate generation, $n$-best reranking, minimum Bayes risk (MBR) decoding, and self-debugging as the core components. We then study the contributions of these components through execution-based evaluation metrics. Our findings highlight the importance of execution-based methods and the difference gap between execution-based and execution-free methods. Furthermore, we assess the impact of filtering based on trial unit tests, a simple and effective strategy that has been often overlooked in prior works. We also propose self-debugging on multiple candidates, obtaining state-of-the-art performance on reranking for code generation. We expect our framework to provide a solid guideline for future research on code generation.
Large Language Models (LLMs) have seen increasing use in various software development tasks, especially in code generation. The most advanced recent methods attempt to incorporate feedback from code execution into prompts to help guide LLMs in generating correct code, in an iterative process. While effective, these methods could be costly and time-consuming due to numerous interactions with the LLM and the extensive token usage. To address this issue, we propose an alternative approach named Evolutionary Prompt Engineering for Code (EPiC), which leverages a lightweight evolutionary algorithm to evolve the original prompts toward better ones that produce high-quality code, with minimal interactions with LLM. Our evaluation against state-of-the-art (SOTA) LLM-based code generation models shows that EPiC outperforms all the baselines in terms of cost-effectiveness.
Recent advancements in large language models (LLMs) have showcased impressive code generation capabilities, primarily evaluated through language-to-code benchmarks. However, these benchmarks may not fully capture a model's code understanding abilities. We introduce CodeJudge-Eval (CJ-Eval), a novel benchmark designed to assess LLMs' code understanding abilities from the perspective of code judging rather than code generation. CJ-Eval challenges models to determine the correctness of provided code solutions, encompassing various error types and compilation issues. By leveraging a diverse set of problems and a fine-grained judging system, CJ-Eval addresses the limitations of traditional benchmarks, including the potential memorization of solutions. Evaluation of 12 well-known LLMs on CJ-Eval reveals that even state-of-the-art models struggle, highlighting the benchmark's ability to probe deeper into models' code understanding abilities. Our codes and benchmark are available at \url{https://github.com/CodeLLM-Research/CodeJudge-Eval}.
The application of large-language models (LLMs) to digital hardware code generation is an emerging field, with most LLMs primarily trained on natural language and software code. Hardware code like Verilog constitutes a small portion of training data, and few hardware benchmarks exist. The open-source VerilogEval benchmark, released in November 2023, provided a consistent evaluation framework for LLMs on code completion tasks. Since then, both commercial and open models have seen significant development. In this work, we evaluate new commercial and open models since VerilogEval's original release-including GPT-4o, GPT-4 Turbo, Llama3.1 (8B/70B/405B), Llama3 70B, Mistral Large, DeepSeek Coder (33B and 6.7B), CodeGemma 7B, and RTL-Coder-against an improved VerilogEval benchmark suite. We find measurable improvements in state-of-the-art models: GPT-4o achieves a 63% pass rate on specification-to-RTL tasks. The recently released and open Llama3.1 405B achieves a 58% pass rate, almost matching GPT-4o, while the smaller domain-specific RTL-Coder 6.7B models achieve an impressive 34% pass rate. Additionally, we enhance VerilogEval's infrastructure by automatically classifying failures, introducing in-context learning support, and extending the tasks to specification-to-RTL translation. We find that prompt engineering remains crucial for achieving good pass rates and varies widely with model and task. A benchmark infrastructure that allows for prompt engineering and failure analysis is essential for continued model development and deployment.
Recent advances in large language models (LLMs) have transformed software development by automatically generating code from natural language. Yet challenges remain in generating fully correct code that aligns with user intent. Our study reveals that LLMs tend to pay less attention to user prompts as more code tokens are generated. We hypothesize that this attention dilution issue is an important reason for code generation errors. To mitigate this issue, we propose Selective Prompt Anchoring (SPA) to guide code LLMs to pay more attention to user intent when generating code. We evaluate SPA using six base LLMs across six benchmarks. Our results demonstrate that SPA enhances Pass@1 by up to 12.9%, consistently outperforming SOTA code generation methods in all settings. Our code is available at https://github.com/magic-YuanTian/Selective-Prompt-Anchoring.