Existing methods fail to effectively steer Large Language Models (LLMs) between textual reasoning and code generation, leaving symbolic computing capabilities underutilized. We introduce CodeSteer, an effective method for guiding LLM code/text generation. We construct a comprehensive benchmark SymBench comprising 37 symbolic tasks with adjustable complexity and also synthesize datasets of 12k multi-turn guidance/generation trajectories and 5.5k guidance comparison pairs. We fine-tune the Llama-3-8B model with a newly designed multi-turn supervised fine-tuning (SFT) and direct preference optimization (DPO). The resulting model, CodeSteerLLM, augmented with the proposed symbolic and self-answer checkers, effectively guides the code/text generation of larger models. Augmenting GPT-4o with CodeSteer raises its average performance score from 53.3 to 86.4, even outperforming the existing best LLM OpenAI o1 (82.7), o1-preview (74.8), and DeepSeek R1 (76.8) across all 37 tasks (28 seen, 9 unseen). Trained for GPT-4o, CodeSteer demonstrates superior generalizability, providing an average 41.8 performance boost on Claude, Mistral, and GPT-3.5. CodeSteer-guided LLMs fully harness symbolic computing to maintain strong performance on highly complex tasks. Models, Datasets, and Codes are available at https://github.com/yongchao98/CodeSteer-v1.0 and https://huggingface.co/yongchao98.
Long-form generation is crucial for academic writing papers and repo-level code generation. Despite this, current models, including GPT-4o, still exhibit unsatisfactory performance. Existing methods that utilize preference learning with outcome supervision often fail to provide detailed feedback for extended contexts. This shortcoming can lead to content that does not fully satisfy query requirements, resulting in issues like length deviations, and diminished quality. In this paper, we propose enhancing long-form generation by incorporating process supervision. We employ Monte Carlo Tree Search to gather stepwise preference pairs, utilizing a global memory pool to maintain consistency. To address the issue of suboptimal candidate selection, we integrate external critiques to refine and improve the quality of the preference pairs. Finally, we apply step-level DPO using the collected stepwise preference pairs. Experimental results show that our method improves length and quality on long-form generation benchmarks, with almost lossless performance on general benchmarks across various model backbones.
Since time immemorial an old adage has always seemed to ring true: you cannot use a high-level productive programming language like Python or R for real-time control and embedded-systems programming, you must rewrite your program in C. We present a counterexample to this mantra by demonstrating how recent compiler developments in the Julia programming language allow users of Julia and the equation-based modeling language ModelingToolkit to compile and deploy binaries for real-time model-based estimation and control. Contrary to the approach taken by a majority of modeling and simulation tools, we do not generate C code, and instead demonstrate how we may use the native Julia code-generation pipeline through LLVM to compile architecture-specific binaries from high-level code. This approach avoids many of the restrictions typically placed on high-level languages to enable C-code generation. As case studies, we include a nonlinear state estimator derived from an equation-based model which is compiled into a program that performs state estimation for deployment onto a Raspberry Pi, as well as a PID controller library implemented in Julia and compiled into a shared library callable from a C program.
Large language models based Multi Agent Systems (MAS) have demonstrated promising performance for enhancing the efficiency and accuracy of code generation tasks. However,most existing methods follow a conventional sequence of planning, coding, and debugging,which contradicts the growth-driven nature of human learning process. Additionally,the frequent information interaction between multiple agents inevitably involves high computational costs. In this paper,we propose Cogito,a neurobiologically inspired multi-agent framework to enhance the problem-solving capabilities in code generation tasks with lower cost. Specifically,Cogito adopts a reverse sequence: it first undergoes debugging, then coding,and finally planning. This approach mimics human learning and development,where knowledge is acquired progressively. Accordingly,a hippocampus-like memory module with different functions is designed to work with the pipeline to provide quick retrieval in similar tasks. Through this growth-based learning model,Cogito accumulates knowledge and cognitive skills at each stage,ultimately forming a Super Role an all capable agent to perform the code generation task. Extensive experiments against representative baselines demonstrate the superior performance and efficiency of Cogito. The code is publicly available at https://anonymous.4open.science/r/Cogito-0083.
The irruption of DeepSeek-R1 constitutes a turning point for the AI industry in general and the LLMs in particular. Its capabilities have demonstrated outstanding performance in several tasks, including creative thinking, code generation, maths and automated program repair, at apparently lower execution cost. However, LLMs must adhere to an important qualitative property, i.e., their alignment with safety and human values. A clear competitor of DeepSeek-R1 is its American counterpart, OpenAI's o3-mini model, which is expected to set high standards in terms of performance, safety and cost. In this technical report, we systematically assess the safety level of both DeepSeek-R1 (70b version) and OpenAI's o3-mini (beta version). To this end, we make use of our recently released automated safety testing tool, named ASTRAL. By leveraging this tool, we automatically and systematically generated and executed 1,260 test inputs on both models. After conducting a semi-automated assessment of the outcomes provided by both LLMs, the results indicate that DeepSeek-R1 produces significantly more unsafe responses (12%) than OpenAI's o3-mini (1.2%).
The Handbook of Combinatorial Designs catalogs many types of combinatorial designs, together with lists of open instances for which existence has not yet been determined. We develop a constructive protocol CPro1, which uses Large Language Models (LLMs) to generate code that constructs combinatorial designs and resolves some of these open instances. The protocol starts from a definition of a particular type of design, and a verifier that reliably confirms whether a proposed design is valid. The LLM selects strategies and implements them in code, and scaffolding provides automated hyperparameter tuning and execution feedback using the verifier. Most generated code fails, but by generating many candidates, the protocol automates exploration of a variety of standard methods (e.g. simulated annealing, genetic algorithms) and experimentation with variations (e.g. cost functions) to find successful approaches. Testing on 16 different types of designs, CPro1 constructs solutions to open instances for 6 of them: Symmetric and Skew Weighing Matrices, Equidistant Permutation Arrays, Packing Arrays, Balanced Ternary Designs, and Florentine Rectangles.
Computer programming textbooks and software documentations often contain flowcharts to illustrate the flow of an algorithm or procedure. Modern OCR engines often tag these flowcharts as graphics and ignore them in further processing. In this paper, we work towards making flowchart images machine-interpretable by converting them to executable Python codes. To this end, inspired by the recent success in natural language to code generation literature, we present a novel transformer-based framework, namely FloCo-T5. Our model is well-suited for this task,as it can effectively learn semantics, structure, and patterns of programming languages, which it leverages to generate syntactically correct code. We also used a task-specific pre-training objective to pre-train FloCo-T5 using a large number of logic-preserving augmented code samples. Further, to perform a rigorous study of this problem, we introduce theFloCo dataset that contains 11,884 flowchart images and their corresponding Python codes. Our experiments show promising results, and FloCo-T5 clearly outperforms related competitive baselines on code generation metrics. We make our dataset and implementation publicly available.
Large Language Models (LLMs) have shown impressive performance across a wide array of tasks involving both structured and unstructured textual data. Recent results on various benchmarks for code generation, repair, or completion suggest that certain models have programming abilities comparable to or even surpass humans. In this work, we demonstrate that high performance on such benchmarks does not correlate to humans' innate ability to understand structural control flow in code. To this end, we extract solutions from the HumanEval benchmark, which the relevant models perform strongly on, and trace their execution path using function calls sampled from the respective test set. Using this dataset, we investigate the ability of seven state-of-the-art LLMs to match the execution trace and find that, despite their ability to generate semantically identical code, they possess limited ability to trace execution paths, especially for longer traces and specific control structures. We find that even the top-performing model, Gemini, can fully and correctly generate only 47% of HumanEval task traces. Additionally, we introduce a subset for three key structures not contained in HumanEval: Recursion, Parallel Processing, and Object-Oriented Programming, including concepts like Inheritance and Polymorphism. Besides OOP, we show that none of the investigated models achieve an accuracy over 5% on the relevant traces. Aggregating these specialized parts with HumanEval tasks, we present CoCoNUT: Code Control Flow for Navigation Understanding and Testing, which measures a model's ability to trace execution of code upon relevant calls, including advanced structural components. We conclude that current LLMs need significant improvement to enhance code reasoning abilities. We hope our dataset helps researchers bridge this gap.
Ensuring the correctness of code generated by Large Language Models (LLMs) presents a significant challenge in AI-driven software development. Existing approaches predominantly rely on black-box (closed-box) approaches that evaluate correctness post-generation, failing to utilize the rich insights embedded in the LLMs' internal states during code generation. In this paper, we introduce OPENIA, a novel white-box (open-box) framework that leverages these internal representations to assess the correctness of LLM-generated code. OPENIA systematically analyzes the intermediate states of representative open-source LLMs specialized for code, including DeepSeek-Coder, CodeLlama, and MagicCoder, across diverse code generation benchmarks. Our empirical analysis reveals that these internal representations encode latent information, which strongly correlates with the correctness of the generated code. Building on these insights, OPENIA uses a white-box/open-box approach to make informed predictions about code correctness, offering significant advantages in adaptability and robustness over traditional classification-based methods and zero-shot approaches. Experimental results demonstrate that OPENIA consistently outperforms baseline models, achieving higher accuracy, precision, recall, and F1-Scores with up to a 2X improvement in standalone code generation and a 46% enhancement in repository-specific scenarios. By unlocking the potential of in-process signals, OPENIA paves the way for more proactive and efficient quality assurance mechanisms in LLM-assisted code generation.
Chaos Engineering (CE) is an engineering technique aimed at improving the resiliency of distributed systems. It involves artificially injecting specific failures into a distributed system and observing its behavior in response. Based on the observation, the system can be proactively improved to handle those failures. Recent CE tools implement the automated execution of predefined CE experiments. However, defining these experiments and improving the system based on the experimental results still remain manual. To reduce the costs of the manual operations, we propose ChaosEater, a system for automating the entire CE operations with Large Language Models (LLMs). It predefines the agentic workflow according to a systematic CE cycle and assigns subdivided operations within the workflow to LLMs. ChaosEater targets CE for Kubernetes systems, which are managed through code (i.e., Infrastructure as Code). Therefore, the LLMs in ChaosEater perform software engineering tasks to complete CE cycles, including requirement definition, code generation, debugging, and testing. We evaluate ChaosEater through case studies on both small and large Kubernetes systems. The results demonstrate that it stably completes reasonable single CE cycles with significantly low time and monetary costs. The CE cycles are also qualitatively validated by human engineers and LLMs.
Large Language Models (LLMs) are becoming integral to daily life, showcasing their vast potential across various Natural Language Processing (NLP) tasks. Beyond NLP, LLMs are increasingly used in software development tasks, such as code completion, modification, bug fixing, and code translation. Software engineers widely use tools like GitHub Copilot and Amazon Q, streamlining workflows and automating tasks with high accuracy. While the resource and energy intensity of LLM training is often highlighted, inference can be even more resource-intensive over time, as it's a continuous process with a high number of invocations. Therefore, developing resource-efficient alternatives for LLM inference is crucial for sustainability. This work proposes GREEN-CODE, a framework for energy-aware code generation in LLMs. GREEN-CODE performs dynamic early exit during LLM inference. We train a Reinforcement Learning (RL) agent that learns to balance the trade-offs between accuracy, latency, and energy consumption. Our approach is evaluated on two open-source LLMs, Llama 3.2 3B and OPT 2.7B, using the JavaCorpus and PY150 datasets. Results show that our method reduces the energy consumption between 23-50 % on average for code generation tasks without significantly affecting accuracy.
Optimization models have been applied to solve a wide variety of decision-making problems. These models are usually developed by optimization experts but are used by practitioners without optimization expertise in various application domains. As a result, practitioners often struggle to interact with and draw useful conclusions from optimization models independently. To fill this gap, we introduce OptiChat, a natural language dialogue system designed to help practitioners interpret model formulation, diagnose infeasibility, analyze sensitivity, retrieve information, evaluate modifications, and provide counterfactual explanations. By augmenting large language models (LLMs) with functional calls and code generation tailored for optimization models, we enable seamless interaction and minimize the risk of hallucinations in OptiChat. We develop a new dataset to evaluate OptiChat's performance in explaining optimization models. Experiments demonstrate that OptiChat effectively bridges the gap between optimization models and practitioners, delivering autonomous, accurate, and instant responses.
Large Language Models (LLMs) have significantly aided developers by generating or assisting in code writing, enhancing productivity across various tasks. While identifying incorrect code is often straightforward, detecting vulnerabilities in functionally correct code is more challenging, especially for developers with limited security knowledge, which poses considerable security risks of using LLM-generated code and underscores the need for robust evaluation benchmarks that assess both functional correctness and security. Current benchmarks like CyberSecEval and SecurityEval attempt to solve it but are hindered by unclear and impractical specifications, failing to assess both functionality and security accurately. To tackle these deficiencies, we introduce CWEval, a novel outcome-driven evaluation framework designed to enhance the evaluation of secure code generation by LLMs. This framework not only assesses code functionality but also its security simultaneously with high-quality task specifications and outcome-driven test oracles which provides high accuracy. Coupled with CWEval-bench, a multilingual, security-critical coding benchmark, CWEval provides a rigorous empirical security evaluation on LLM-generated code, overcoming previous benchmarks' shortcomings. Through our evaluations, CWEval reveals a notable portion of functional but insecure code produced by LLMs, and shows a serious inaccuracy of previous evaluations, ultimately contributing significantly to the field of secure code generation. We open-source our artifact at: https://github.com/Co1lin/CWEval .
Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in chart understanding tasks. However, interpreting charts with textual descriptions often leads to information loss, as it fails to fully capture the dense information embedded in charts. In contrast, parsing charts into code provides lossless representations that can effectively contain all critical details. Although existing open-source MLLMs have achieved success in chart understanding tasks, they still face two major challenges when applied to chart-to-code tasks.: (1) Low executability and poor restoration of chart details in the generated code and (2) Lack of large-scale and diverse training data. To address these challenges, we propose \textbf{ChartCoder}, the first dedicated chart-to-code MLLM, which leverages Code LLMs as the language backbone to enhance the executability of the generated code. Furthermore, we introduce \textbf{Chart2Code-160k}, the first large-scale and diverse dataset for chart-to-code generation, and propose the \textbf{Snippet-of-Thought (SoT)} method, which transforms direct chart-to-code generation data into step-by-step generation. Experiments demonstrate that ChartCoder, with only 7B parameters, surpasses existing open-source MLLMs on chart-to-code benchmarks, achieving superior chart restoration and code excitability. Our code will be available at https://github.com/thunlp/ChartCoder.
Large language models (LLMs) have been widely deployed in coding tasks, drawing increasing attention to the evaluation of the quality and safety of LLMs' outputs. However, research on bias in code generation remains limited. Existing studies typically identify bias by applying malicious prompts or reusing tasks and dataset originally designed for discriminative models. Given that prior datasets are not fully optimized for code-related tasks, there is a pressing need for benchmarks specifically designed for evaluating code models. In this study, we introduce FairCoder, a novel benchmark for evaluating social bias in code generation. FairCoder explores the bias issue following the pipeline in software development, from function implementation to unit test, with diverse real-world scenarios. Additionally, three metrics are designed to assess fairness performance on this benchmark. We conduct experiments on widely used LLMs and provide a comprehensive analysis of the results. The findings reveal that all tested LLMs exhibit social bias.
Large reasoning models (LRMs) like OpenAI-o1 have demonstrated impressive long stepwise reasoning capabilities through large-scale reinforcement learning. However, their extended reasoning processes often suffer from knowledge insufficiency, leading to frequent uncertainties and potential errors. To address this limitation, we introduce \textbf{Search-o1}, a framework that enhances LRMs with an agentic retrieval-augmented generation (RAG) mechanism and a Reason-in-Documents module for refining retrieved documents. Search-o1 integrates an agentic search workflow into the reasoning process, enabling dynamic retrieval of external knowledge when LRMs encounter uncertain knowledge points. Additionally, due to the verbose nature of retrieved documents, we design a separate Reason-in-Documents module to deeply analyze the retrieved information before injecting it into the reasoning chain, minimizing noise and preserving coherent reasoning flow. Extensive experiments on complex reasoning tasks in science, mathematics, and coding, as well as six open-domain QA benchmarks, demonstrate the strong performance of Search-o1. This approach enhances the trustworthiness and applicability of LRMs in complex reasoning tasks, paving the way for more reliable and versatile intelligent systems. The code is available at \url{https://github.com/sunnynexus/Search-o1}.
Pre-training a language model and then fine-tuning it has shown to be an efficient and effective technique for a wide range of code intelligence tasks, such as code generation, code summarization, and vulnerability detection. However, pretraining language models on a large-scale code corpus is computationally expensive. Fortunately, many off-the-shelf Pre-trained Code Models (PCMs), such as CodeBERT, CodeT5, CodeGen, and Code Llama, have been released publicly. These models acquire general code understanding and generation capability during pretraining, which enhances their performance on downstream code intelligence tasks. With an increasing number of these public pre-trained models, selecting the most suitable one to reuse for a specific task is essential. In this paper, we systematically investigate the reusability of PCMs. We first explore three intuitive model selection methods that select by size, training data, or brute-force fine-tuning. Experimental results show that these straightforward techniques either perform poorly or suffer high costs. Motivated by these findings, we explore learning-based model selection strategies that utilize pre-trained models without altering their parameters. Specifically, we train proxy models to gauge the performance of pre-trained models, and measure the distribution deviation between a model's latent features and the task's labels, using their closeness as an indicator of model transferability. We conduct experiments on 100 widely-used opensource PCMs for code intelligence tasks, with sizes ranging from 42.5 million to 3 billion parameters. The results demonstrate that learning-based selection methods reduce selection time to 100 seconds, compared to 2,700 hours with brute-force fine-tuning, with less than 6% performance degradation across related tasks.
Large Language Models (LLMs) have demonstrated remarkable performance in various NLP tasks, including semantic parsing, which trans lates natural language into formal code representations. However, the reverse process, translating code into natural language, termed semantic captioning, has received less attention. This task is becoming increasingly important as LLMs are integrated into platforms for code generation, security analysis, and educational purposes. In this paper, we focus on the captioning of SQL query (SQL2Text) to address the critical need for understanding and explaining SQL queries in an era where LLM-generated code poses potential security risks. We repurpose Text2SQL datasets for SQL2Text by introducing an iterative ICL prompt using GPT-4o to generate multiple additional utterances, which enhances the robustness of the datasets for the reverse task. We conduct our experiments using in-context learning (ICL) based on different sample selection methods, emphasizing smaller, more computationally efficient LLMs. Our findings demonstrate that leveraging the inherent graph properties of SQL for ICL sample selection significantly outperforms random selection by up to 39% on BLEU score and provides better results than alternative methods. Dataset and codes are published: \url{https://github.com/aliwister/ast-icl}.
As large language models (LLMs) are increasingly deployed in diverse applications, including chatbot assistants and code generation, aligning their behavior with safety and ethical standards has become paramount. However, jailbreak attacks, which exploit vulnerabilities to elicit unintended or harmful outputs, threaten LLMs' safety significantly. In this paper, we introduce Layer-AdvPatcher, a novel methodology designed to defend against jailbreak attacks by utilizing an unlearning strategy to patch specific layers within LLMs through self-augmented datasets. Our insight is that certain layer(s), tend to produce affirmative tokens when faced with harmful prompts. By identifying these layers and adversarially exposing them to generate more harmful data, one can understand their inherent and diverse vulnerabilities to attacks. With these exposures, we then "unlearn" these issues, reducing the impact of affirmative tokens and hence minimizing jailbreak risks while keeping the model's responses to safe queries intact. We conduct extensive experiments on two models, four benchmark datasets, and multiple state-of-the-art jailbreak attacks to demonstrate the efficacy of our approach. Results indicate that our framework reduces the harmfulness and attack success rate of jailbreak attacks without compromising utility for benign queries compared to recent defense methods. Our code is publicly available at: https://github.com/oyy2000/LayerAdvPatcher
This bachelor's thesis examines the capabilities of ChatGPT 4 in code generation across 19 programming languages. The study analyzed solution rates across three difficulty levels, types of errors encountered, and code quality in terms of runtime and memory efficiency through a quantitative experiment. A total of 188 programming problems were selected from the LeetCode platform, and ChatGPT 4 was given three attempts to produce a correct solution with feedback. ChatGPT 4 successfully solved 39.67% of all tasks, with success rates decreasing significantly as problem complexity increased. Notably, the model faced considerable challenges with hard problems across all languages. ChatGPT 4 demonstrated higher competence in widely used languages, likely due to a larger volume and higher quality of training data. The solution rates also revealed a preference for languages with low abstraction levels and static typing. For popular languages, the most frequent error was "Wrong Answer," whereas for less popular languages, compiler and runtime errors prevailed, suggesting frequent misunderstandings and confusion regarding the structural characteristics of these languages. The model exhibited above-average runtime efficiency in all programming languages, showing a tendency toward statically typed and low-abstraction languages. Memory efficiency results varied significantly, with above-average performance in 14 languages and below-average performance in five languages. A slight preference for low-abstraction languages and a leaning toward dynamically typed languages in terms of memory efficiency were observed. Future research should include a larger number of tasks, iterations, and less popular languages. Additionally, ChatGPT 4's abilities in code interpretation and summarization, debugging, and the development of complex, practical code could be analyzed further.