arXiv subjects

Zhi Jin

Publications and source records attributed to Zhi Jin.

At least 127 records · Page 7Linked to original sources

arXiv2024

Showing LLM-Generated Code Selectively Based on Confidence of LLMs

Large Language Models (LLMs) have shown impressive abilities in code generation, but they may generate erroneous programs. Reading a program takes ten times longer than writing it. Showing these erroneous programs to developers will waste developers' energies and introduce security risks to software. To address the above limitations, we propose HonestCoder, a novel LLM-based code generation approach. HonestCoder selectively shows the generated programs to developers based on LLMs' confidence. The confidence provides valuable insights into the correctness of generated programs. To achieve this goal, we propose a novel approach to estimate LLMs' confidence in code generation. It estimates confidence by measuring the multi-modal similarity between LLMs-generated programs. We collect and release a multilingual benchmark named TruthCodeBench, which consists of 2,265 samples and covers two popular programming languages (i.e., Python and Java). We apply HonestCoder to four popular LLMs (e.g., DeepSeek-Coder and Code Llama) and evaluate it on TruthCodeBench. Based on the experiments, we obtain the following insights. (1) HonestCoder can effectively estimate LLMs' confidence and accurately determine the correctness of generated programs. For example, HonestCoder outperforms the state-of-the-art baseline by 27.79% in AUROC and 63.74% in AUCPR. (2) HonestCoder can decrease the number of erroneous programs shown to developers. Compared to eight baselines, it can show more correct programs and fewer erroneous programs to developers. (3) Compared to showing code indiscriminately, HonestCoder only adds slight time overhead (approximately 0.4 seconds per requirement). (4) We discuss future directions to facilitate the application of LLMs in software development. We hope this work can motivate broad discussions about measuring the reliability of LLMs' outputs in performing code-related tasks.

cs.SE↗

arXiv2024

Generating Equivalent Representations of Code By A Self-Reflection Approach

Equivalent Representations (ERs) of code are textual representations that preserve the same semantics as the code itself, e.g., natural language comments and pseudocode. ERs play a critical role in software development and maintenance. However, how to automatically generate ERs of code remains an open challenge. In this paper, we propose a self-reflection approach to generating ERs of code. It enables two Large Language Models (LLMs) to work mutually and produce an ER through a reflection process. Depending on whether constraints on ERs are applied, our approach generates ERs in both open and constrained settings. We conduct a empirical study to generate ERs in two settings and obtain eight findings. (1) Generating ERs in the open setting. In the open setting, we allow LLMs to represent code without any constraints, analyzing the resulting ERs and uncovering five key findings. These findings shed light on how LLMs comprehend syntactic structures, APIs, and numerical computations in code. (2) Generating ERs in the constrained setting. In the constrained setting, we impose constraints on ERs, such as natural language comments, pseudocode, and flowcharts. This allows our approach to address a range of software engineering tasks. Based on our experiments, we have three findings demonstrating that our approach can effectively generate ERs that adhere to specific constraints, thus supporting various software engineering tasks. (3) Future directions. We also discuss potential future research directions, such as deriving intermediate languages for code generation, exploring LLM-friendly requirement descriptions, and further supporting software engineering tasks. We believe that this paper will spark discussions in research communities and inspire many follow-up studies.

cs.CL↗

arXiv2024

NTIRE 2024 Challenge on Stereo Image Super-Resolution: Methods and Results

This paper summarizes the 3rd NTIRE challenge on stereo image super-resolution (SR) with a focus on new solutions and results. The task of this challenge is to super-resolve a low-resolution stereo image pair to a high-resolution one with a magnification factor of x4 under a limited computational budget. Compared with single image SR, the major challenge of this challenge lies in how to exploit additional information in another viewpoint and how to maintain stereo consistency in the results. This challenge has 2 tracks, including one track on bicubic degradation and one track on real degradations. In total, 108 and 70 participants were successfully registered for each track, respectively. In the test phase, 14 and 13 teams successfully submitted valid results with PSNR (RGB) scores better than the baseline. This challenge establishes a new benchmark for stereo image SR.

cs.CV↗

arXiv2024

Sifting through the Chaff: On Utilizing Execution Feedback for Ranking the Generated Code Candidates

Large Language Models (LLMs), such as GPT-4, StarCoder, and CodeLlama, are transforming the way developers approach programming by automatically generating code based on given natural language descriptions. Despite advancements, generating syntactically and semantically correct code remains challenging, especially for complex programming tasks. Existing approaches typically generate multiple candidate solutions using LLMs to increase the likelihood of producing correct code. However, selecting the correct code from these candidates-a process known as code ranking-remains a major challenge. Current research on code ranking can be categorized into execution-based and non-execution-based methods. Execution-based methods, although effective, encounter notable limitations, such as scarcity of quality unit tests and security risks. Non-execution-based methods like CodeRanker, which rely solely on classification labels to train a code ranker, struggle to capture subtle errors and provide detailed error insights. Recognizing the strengths and limitations of both approaches, we propose a new method. The key insight of our work is that an effective code ranker is expected to truly comprehend the underlying causes of erroneous code, as relying solely on classification labels is insufficient. Inspired by this, this paper puts forward RankEF, an innovative approach for code ranking that leverages execution feedback. RankEF employs multi-task learning to integrate code classification with execution feedback generation. This approach enables the model to understand the reasons behind incorrect code, distinguishing between correct and incorrect solutions without the need to execute the code during the ranking phase. Experiments on three code generation benchmarks demonstrate that RankEF significantly outperforms the state-of-the-art CodeRanker.

cs.SE↗

arXiv2024

Python Symbolic Execution with LLM-powered Code Generation

Symbolic execution is a key technology in software testing, which generates test cases by collecting symbolic path constraints and then solving constraints with SMT solvers. Symbolic execution has been proven helpful in generating high-coverage test cases, but its limitations, e.g., the difficulties in solving path constraints, prevent it from broader usage in software testing. Moreover, symbolic execution has encountered many difficulties when applied to dynamically typed languages like Python, because it is extremely challenging to translate the flexible Python grammar into rigid solvers. To overcome the main challenges of applying symbolic execution in Python, we proposed an LLM-empowered agent, LLM-Sym, that automatically calls an SMT solver, Z3, to solve execution path constraints. Based on an introductory-level symbolic execution engine, our LLM agent can extend it to supporting programs with complex data type `list'. The core contribution of LLM-Sym is translating complex Python path constraints into Z3 code. To enable accurate path-to-Z3 translation, we design a multiple-step code generation pipeline including type inference, retrieval and self-refine. Our experiments demonstrate that LLM-Sym is capable of solving path constraints on Leetcode problems with complicated control flows and list data structures, which is impossible for the backbone symbolic execution engine. Our approach paves the way for the combination of the generation ability of LLMs with the reasoning ability of symbolic solvers, and opens up new opportunities in LLM-augmented test case generation.

cs.SE↗

arXiv2024

GraphCoder: Enhancing Repository-Level Code Completion via Code Context Graph-based Retrieval and Language Model

The performance of repository-level code completion depends upon the effective leverage of both general and repository-specific knowledge. Despite the impressive capability of code LLMs in general code completion tasks, they often exhibit less satisfactory performance on repository-level completion due to the lack of repository-specific knowledge in these LLMs. To address this problem, we propose GraphCoder, a retrieval-augmented code completion framework that leverages LLMs' general code knowledge and the repository-specific knowledge via a graph-based retrieval-generation process. In particular, GraphCoder captures the context of completion target more accurately through code context graph (CCG) that consists of control-flow, data- and control-dependence between code statements, a more structured way to capture the completion target context than the sequence-based context used in existing retrieval-augmented approaches; based on CCG, GraphCoder further employs a coarse-to-fine retrieval process to locate context-similar code snippets with the completion target from the current repository. Experimental results demonstrate both the effectiveness and efficiency of GraphCoder: Compared to baseline retrieval-augmented methods, GraphCoder achieves higher exact match (EM) on average, with increases of +6.06 in code match and +6.23 in identifier match, while using less time and space.

cs.SE↗

arXiv2024

CodeScore: Evaluating Code Generation by Learning Code Execution

A proper code evaluation metric (CEM) profoundly impacts the evolution of code generation, which is an important research field in NLP and software engineering. Prevailing match-based CEMs (e.g., BLEU, Accuracy, and CodeBLEU) suffer from two significant drawbacks. 1. They primarily measure the surface differences between codes without considering their functional equivalence. However, functional equivalence is pivotal in evaluating the effectiveness of code generation, as different codes can perform identical operations. 2. They are predominantly designed for the Ref-only input format. However, code evaluation necessitates versatility in input formats. Aside from Ref-only, there are NL-only and Ref\&NL formats, which existing match-based CEMs cannot effectively accommodate. In this paper, we propose CodeScore, a large language model (LLM)-based CEM, which estimates the functional correctness of generated code on three input types. To acquire CodeScore, we present UniCE, a unified code generation learning framework, for LLMs to learn code execution (i.e., learning PassRatio and Executability of generated code) with unified input. Extensive experimental results on multiple code evaluation datasets demonstrate that CodeScore absolutely improves up to 58.87% correlation with functional correctness compared to other CEMs, achieves state-of-the-art performance, and effectively handles three input formats.

cs.SE↗

arXiv2024

HITS: High-coverage LLM-based Unit Test Generation via Method Slicing

Large language models (LLMs) have behaved well in generating unit tests for Java projects. However, the performance for covering the complex focal methods within the projects is poor. Complex methods comprise many conditions and loops, requiring the test cases to be various enough to cover all lines and branches. However, existing test generation methods with LLMs provide the whole method-to-test to the LLM without assistance on input analysis. The LLM has difficulty inferring the test inputs to cover all conditions, resulting in missing lines and branches. To tackle the problem, we propose decomposing the focal methods into slices and asking the LLM to generate test cases slice by slice. Our method simplifies the analysis scope, making it easier for the LLM to cover more lines and branches in each slice. We build a dataset comprising complex focal methods collected from the projects used by existing state-of-the-art approaches. Our experiment results show that our method significantly outperforms current test case generation methods with LLMs and the typical SBST method Evosuite regarding both line and branch coverage scores.

cs.SE↗

arXiv2024

CodeAgent: Enhancing Code Generation with Tool-Integrated Agent Systems for Real-World Repo-level Coding Challenges

Large Language Models (LLMs) have shown promise in automated code generation but typically excel only in simpler tasks such as generating standalone code units. Real-world software development, however, often involves complex code repositories (named repo) with complex dependencies and extensive documentation. To fill this gap, our research pivots towards evaluating LLMs in a more realistic setting -- real-world repo-level code generation. We introduce CodeAgentBench, a manually curated benchmark for repo-level code generation. This benchmark comprises five high-quality Python projects, encompassing a total of 101 samples. We assess nine leading LLMs on repo-level tasks and observe a decline in their performance. To tackle this, we present CodeAgent, a novel LLM-based agent framework that employs external tools for effective repo-level code generation. CodeAgent integrates five programming tools, enabling interaction with software artifacts for information retrieval, code symbol navigation, and code testing. We implement four agent strategies to optimize these tools' usage. Our experiments on CodeAgentBench show that CodeAgent enhances LLM performance significantly, with improvements ranging from 18.1\% to 250\%. Further tests on the HumanEval benchmark confirm CodeAgent's adaptability and efficacy across various code generation tasks. Notably, CodeAgent outperforms commercial products like Github Copilot, showcasing superior accuracy and efficiency. These results demonstrate CodeAgent's robust capabilities in code generation, highlighting its potential for real-world repo-level coding challenges.

cs.SE↗

arXiv2024

HiRoPE: Length Extrapolation for Code Models Using Hierarchical Position

Addressing the limitation of context length in large language models for code-related tasks is the primary focus of this paper. Existing LLMs are constrained by their pre-trained context lengths, leading to performance issues in handling long complex code sequences. Inspired by how human programmers navigate code, we introduce Hierarchical Rotary Position Embedding (HiRoPE), a novel approach that enhances the traditional rotary position embedding into a hierarchical format based on the hierarchical structure of source code. HiRoPE offers easy integration into existing LLMs without extra training costs. Our method is extensively evaluated with various LLMs, demonstrating stable performance in tasks such as language modeling and long code completion. We also introduce a new long code understanding task with real-world code projects, in hopes of promoting further development in this code-related field. Theoretically and experimentally, we find that HiRoPE also addresses the out-of-distribution issue in position encoding. Our HiRoPE significantly expands the context length capabilities of LLMs, enabling inference at lengths exponentially greater than the training length.

cs.SE↗

arXiv2024

An Evaluation of Requirements Modeling for Cyber-Physical Systems via LLMs

Cyber-physical systems (CPSs) integrate cyber and physical components and enable them to interact with each other to meet user needs. The needs for CPSs span rich application domains such as healthcare and medicine, smart home, smart building, etc. This indicates that CPSs are all about solving real-world problems. With the increasing abundance of sensing devices and effectors, the problems wanted to solve with CPSs are becoming more and more complex. It is also becoming increasingly difficult to extract and express CPS requirements accurately. Problem frame approach aims to shape real-world problems by capturing the characteristics and interconnections of components, where the problem diagram is central to expressing the requirements. CPSs requirements are generally presented in domain-specific documents that are normally expressed in natural language. There is currently no effective way to extract problem diagrams from natural language documents. CPSs requirements extraction and modeling are generally done manually, which is time-consuming, labor-intensive, and error-prone. Large language models (LLMs) have shown excellent performance in natural language understanding. It can be interesting to explore the abilities of LLMs to understand domain-specific documents and identify modeling elements, which this paper is working on. To achieve this goal, we first formulate two tasks (i.e., entity recognition and interaction extraction) and propose a benchmark called CPSBench. Based on this benchmark, extensive experiments are conducted to evaluate the abilities and limitations of seven advanced LLMs. We find some interesting insights. Finally, we establish a taxonomy of LLMs hallucinations in CPSs requirements modeling using problem diagrams. These results will inspire research on the use of LLMs for automated CPSs requirements modeling.

cs.SE↗

arXiv2024

Integrating Physician Diagnostic Logic into Large Language Models: Preference Learning from Process Feedback

The use of large language models in medical dialogue generation has garnered significant attention, with a focus on improving response quality and fluency. While previous studies have made progress in optimizing model performance for single-round medical Q&A tasks, there is a need to enhance the model's capability for multi-round conversations to avoid logical inconsistencies. To address this, we propose an approach called preference learning from process feedback~(PLPF), which integrates the doctor's diagnostic logic into LLMs. PLPF involves rule modeling, preference data generation, and preference alignment to train the model to adhere to the diagnostic process. Experimental results using Standardized Patient Testing show that PLPF enhances the diagnostic accuracy of the baseline model in medical conversations by 17.6%, outperforming traditional reinforcement learning from human feedback. Additionally, PLPF demonstrates effectiveness in both multi-round and single-round dialogue tasks, showcasing its potential for improving medical dialogue generation.

cs.CL↗

arXiv2024

A Comprehensive Evaluation on Event Reasoning of Large Language Models

Event reasoning is a fundamental ability that underlies many applications. It requires event schema knowledge to perform global reasoning and needs to deal with the diversity of the inter-event relations and the reasoning paradigms. How well LLMs accomplish event reasoning on various relations and reasoning paradigms remains unknown. To mitigate this disparity, we comprehensively evaluate the abilities of event reasoning of LLMs. We introduce a novel benchmark EV2 for EValuation of EVent reasoning. EV2 consists of two levels of evaluation of schema and instance and is comprehensive in relations and reasoning paradigms. We conduct extensive experiments on EV2. We find that LLMs have abilities to accomplish event reasoning but their performances are far from satisfactory. We also notice the imbalance of event reasoning abilities in LLMs. Besides, LLMs have event schema knowledge, however, they're not aligned with humans on how to utilize the knowledge. Based on these findings, we guide the LLMs in utilizing the event schema knowledge as memory leading to improvements on event reasoning.

cs.CL↗

arXiv2024

Accelerating Learned Video Compression via Low-Resolution Representation Learning

In recent years, the field of learned video compression has witnessed rapid advancement, exemplified by the latest neural video codecs DCVC-DC that has outperformed the upcoming next-generation codec ECM in terms of compression ratio. Despite this, learned video compression frameworks often exhibit low encoding and decoding speeds primarily due to their increased computational complexity and unnecessary high-resolution spatial operations, which hugely hinder their applications in reality. In this work, we introduce an efficiency-optimized framework for learned video compression that focuses on low-resolution representation learning, aiming to significantly enhance the encoding and decoding speeds. Firstly, we diminish the computational load by reducing the resolution of inter-frame propagated features obtained from reused features of decoded frames, including I-frames. We implement a joint training strategy for both the I-frame and P-frame models, further improving the compression ratio. Secondly, our approach efficiently leverages multi-frame priors for parameter prediction, minimizing computation at the decoding end. Thirdly, we revisit the application of the Online Encoder Update (OEU) strategy for high-resolution sequences, achieving notable improvements in compression ratio without compromising decoding efficiency. Our efficiency-optimized framework has significantly improved the balance between compression ratio and speed for learned video compression. In comparison to traditional codecs, our method achieves performance levels on par with the low-decay P configuration of the H.266 reference software VTM. Furthermore, when contrasted with DCVC-HEM, our approach delivers a comparable compression ratio while boosting encoding and decoding speeds by a factor of 3 and 7, respectively. On RTX 2080Ti, our method can decode each 1080p frame under 100ms.

eess.IV↗

arXiv2024

M2CVD: Enhancing Vulnerability Semantic through Multi-Model Collaboration for Code Vulnerability Detection

Large Language Models (LLMs) have strong capabilities in code comprehension, but fine-tuning costs and semantic alignment issues limit their project-specific optimization; conversely, code models such CodeBERT are easy to fine-tune, but it is often difficult to learn vulnerability semantics from complex code languages. To address these challenges, this paper introduces the Multi-Model Collaborative Vulnerability Detection approach (M2CVD) that leverages the strong capability of analyzing vulnerability semantics from LLMs to improve the detection accuracy of code models. M2CVD employs a novel collaborative process: first enhancing the quality of vulnerability semantic description produced by LLMs through the understanding of project code by code models, and then using these improved vulnerability semantic description to boost the detection accuracy of code models. We demonstrated M2CVD's effectiveness on two real-world datasets, where M2CVD significantly outperformed the baseline. In addition, we demonstrate that the M2CVD collaborative method can extend to other different LLMs and code models to improve their accuracy in vulnerability detection tasks.

cs.SE↗

arXiv2024

CAPformer: Compression-Aware Pre-trained Transformer for Low-Light Image Enhancement

Low-Light Image Enhancement (LLIE) has advanced with the surge in phone photography demand, yet many existing methods neglect compression, a crucial concern for resource-constrained phone photography. Most LLIE methods overlook this, hindering their effectiveness. In this study, we investigate the effects of JPEG compression on low-light images and reveal substantial information loss caused by JPEG due to widespread low pixel values in dark areas. Hence, we propose the Compression-Aware Pre-trained Transformer (CAPformer), employing a novel pre-training strategy to learn lossless information from uncompressed low-light images. Additionally, the proposed Brightness-Guided Self-Attention (BGSA) mechanism enhances rational information gathering. Experiments demonstrate the superiority of our approach in mitigating compression effects on LLIE, showcasing its potential for improving LLIE in resource-constrained scenarios.

cs.CV↗

arXiv2024

A Survey on Self-Evolution of Large Language Models

Large language models (LLMs) have significantly advanced in various fields and intelligent agent applications. However, current LLMs that learn from human or external model supervision are costly and may face performance ceilings as task complexity and diversity increase. To address this issue, self-evolution approaches that enable LLM to autonomously acquire, refine, and learn from experiences generated by the model itself are rapidly growing. This new training paradigm inspired by the human experiential learning process offers the potential to scale LLMs towards superintelligence. In this work, we present a comprehensive survey of self-evolution approaches in LLMs. We first propose a conceptual framework for self-evolution and outline the evolving process as iterative cycles composed of four phases: experience acquisition, experience refinement, updating, and evaluation. Second, we categorize the evolution objectives of LLMs and LLM-based agents; then, we summarize the literature and provide taxonomy and insights for each module. Lastly, we pinpoint existing challenges and propose future directions to improve self-evolution frameworks, equipping researchers with critical insights to fast-track the development of self-evolving LLMs. Our corresponding GitHub repository is available at https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/Awesome-Self-Evolution-of-LLM

cs.CL↗

arXiv2024

Generalization or Memorization: Data Contamination and Trustworthy Evaluation for Large Language Models

Recent statements about the impressive capabilities of large language models (LLMs) are usually supported by evaluating on open-access benchmarks. Considering the vast size and wide-ranging sources of LLMs' training data, it could explicitly or implicitly include test data, leading to LLMs being more susceptible to data contamination. However, due to the opacity of training data, the black-box access of models, and the rapid growth of synthetic training data, detecting and mitigating data contamination for LLMs faces significant challenges. In this paper, we propose CDD, which stands for Contamination Detection via output Distribution for LLMs. CDD necessitates only the sampled texts to detect data contamination, by identifying the peakedness of LLM's output distribution. To mitigate the impact of data contamination in evaluation, we also present TED: Trustworthy Evaluation via output Distribution, based on the correction of LLM's output distribution. To facilitate this study, we introduce two benchmarks, i.e., DetCon and ComiEval, for data contamination detection and contamination mitigation evaluation tasks. Extensive experimental results show that CDD achieves the average relative improvements of 21.8\%-30.2\% over other contamination detection approaches in terms of Accuracy, F1 Score, and AUC metrics, and can effectively detect implicit contamination. TED substantially mitigates performance improvements up to 66.9\% attributed to data contamination across various contamination setups. In real-world applications, we reveal that ChatGPT exhibits a high potential to suffer from data contamination on HumanEval benchmark.

cs.CL↗