Yuhao Qing

Qi Hu, Yifeng Tang, Qinghua Wang et al.

Large language models are increasingly deployed as coding agents, shifting safety from individual responses to action sequences. Existing benchmarks, however, primarily assess whether models refuse unsafe prompts, leaving impacts on stateful workspaces largely unexamined. We present SABER, a benchmark for environment-aware operational safety that places models in realistic agent-style projects and evaluates safety from the final environment state after a sequence of actions. Beyond binary safety-violation reports, SABER categorizes violations by cause, enabling analysis of model-specific safety profiles. Our evaluations show that even the best-performing model has more than a 54% harmful safety-violation rate (HSR), suggesting that current alignment remains insufficient for realistic project environments. SABER further reveals distinct safety profiles across models. Our benchmark is publicly available at https://github.com/sssr-lab/saber.

Zongyuan Zhang, Qingwen Bu, Tianyang Duan et al. · cmu

Deep neural networks (DNNs) are vulnerable to adversarial examples, in which DNNs are misled to false outputs due to inputs containing imperceptible perturbations. Adversarial training, a reliable and effective method of defense, may significantly reduce the vulnerability of neural networks and becomes the de facto standard for robust learning. While many recent works practice the data-centric philosophy, such as how to generate better adversarial examples or use generative models to produce additional training data, we look back to the models themselves and revisit the adversarial robustness from the perspective of deep feature distribution as an insightful complementarity. In this paper, we propose \textit{Branch Orthogonality adveRsarial Training} (BORT) to obtain state-of-the-art performance with solely the original dataset for adversarial training. To practice our design idea of integrating multiple orthogonal solution spaces, we leverage a simple multi-branch neural network and propose a corresponding loss function, branch-orthogonal loss, to make each solution space of the multi-branch model orthogonal. We evaluate our approach on CIFAR-10, CIFAR-100 and SVHN against $\ell_{\infty}$ norm-bounded perturbations of size $ε= 8/255$, respectively. Exhaustive experiments are conducted to show that our method goes beyond all state-of-the-art methods without any tricks. Compared to all methods that do not use additional data for training, our models achieve 67.3\% and 41.5\% robust accuracy on CIFAR-10 and CIFAR-100 (improving upon the state-of-the-art by +7.23\% and +9.07\%).

Dong Huang, Qingwen Bu, Yuhao Qing et al.

Chain-of-thought (CoT) has emerged as a groundbreaking tool in NLP, notably for its efficacy in complex reasoning tasks, such as mathematical proofs. However, its application in code generation faces a distinct challenge, i.e., although the code generated with CoT reasoning is logically correct, it faces the problem of syntax error (e.g., invalid syntax error report) during code execution, which causes the CoT result's pass@1 in HumanEval even lower than the zero-shot result. In this paper, we present Code Chain-of-Thought (CodeCoT) that integrates CoT with a self-examination process for code generation. CodeCoT begins with the LLMs using CoT for initial code development to ensure the generated code follows the correct logic flow. Then, CodeCoT will generate test cases to validate whether the code has syntax errors during the execution. CodeCoT then employs a self-examination phase, in which the generated code is executed against these test cases in the local environment. If the local environment raises error information (e.g., invalid syntax error), CodeCoT will iteratively refine the code based on the feedback information. Within this loop, CodeCoT can make sure their generated codes not only follow the logic flow of the code description, but the syntax error will also be addressed with the self-examination process. Our evaluation results reveal that CodeCoT improves the effectiveness of code generation. For example, CodeCoT increases pass@1 from 75.6% to 79.3% for the HumanEval dataset.

Dong Huang, Yuhao Qing, Weiyi Shang et al.

Code generation models have increasingly become integral to aiding software development. Although current research has thoroughly examined the correctness of the code produced by code generation models, a vital aspect that plays a pivotal role in green computing and sustainability efforts has often been neglected. This paper presents EffiBench, a benchmark with 1,000 efficiency-critical coding problems to assess the efficiency of code generated by code generation models. EffiBench contains a diverse set of LeetCode coding problems. Each problem is paired with an executable human-written canonical solution, which obtains the SOTA efficiency on the LeetCode solution leaderboard. With EffiBench, we empirically examine the ability of 42 large language models (35 open-source and 7 closed-source) to generate efficient code. Our evaluation results demonstrate that the efficiency of the code generated by LLMs is generally worse than the efficiency of human-written canonical solutions. For example, GPT-4 generated code has an average \textbf{3.12} times execution time that of the human-written canonical solutions. In the most extreme cases, the execution time and total memory usage of GPT-4 generated code are \textbf{13.89} and \textbf{43.92} times that of the canonical solutions. The source code of EffiBench is released on https://github.com/huangd1999/EffiBench. We also provide the LeaderBoard at https://huggingface.co/spaces/EffiBench/effibench-leaderboard.

Dong Huang, Qingwen Bu, Yichao Fu et al.

Deep Neural Networks~(DNNs) have been widely deployed in software to address various tasks~(e.g., autonomous driving, medical diagnosis). However, they could also produce incorrect behaviors that result in financial losses and even threaten human safety. To reveal the incorrect behaviors in DNN and repair them, DNN developers often collect rich unlabeled datasets from the natural world and label them to test the DNN models. However, properly labeling a large number of unlabeled datasets is a highly expensive and time-consuming task. To address the above-mentioned problem, we propose NSS, Neuron Sensitivity guided test case Selection, which can reduce the labeling time by selecting valuable test cases from unlabeled datasets. NSS leverages the internal neuron's information induced by test cases to select valuable test cases, which have high confidence in causing the model to behave incorrectly. We evaluate NSS with four widely used datasets and four well-designed DNN models compared to SOTA baseline methods. The results show that NSS performs well in assessing the test cases' probability of fault triggering and model improvement capabilities. Specifically, compared with baseline approaches, NSS obtains a higher fault detection rate~(e.g., when selecting 5\% test case from the unlabeled dataset in MNIST \& LeNet1 experiment, NSS can obtain 81.8\% fault detection rate, 20\% higher than baselines).

Yuhao Qing, Boyu Zhu, Mingzhe Du et al. · mit

Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this gap, we introduce EffiBench-X, the first multi-language benchmark designed to measure the efficiency of LLM-generated code. EffiBench-X supports Python, C++, Java, JavaScript, Ruby, and Golang. It comprises competitive programming tasks with human-expert solutions as efficiency baselines. Evaluating state-of-the-art LLMs on EffiBench-X reveals that while models generate functionally correct code, they consistently underperform human experts in efficiency. Even the most efficient LLM-generated solutions (Qwen3-32B) achieve only around \textbf{62\%} of human efficiency on average, with significant language-specific variations. LLMs show better efficiency in Python, Ruby, and JavaScript than in Java, C++, and Golang. For instance, DeepSeek-R1's Python code is significantly more efficient than its Java code. These results highlight the critical need for research into LLM optimization techniques to improve code efficiency across diverse languages. The dataset and evaluation infrastructure are submitted and available at https://github.com/EffiBench/EffiBench-X.git and https://huggingface.co/datasets/EffiBench/effibench-x.

Dong Huang, Guangtao Zeng, Jianbo Dai et al.

As large language models (LLMs) play an increasingly important role in code generation, enhancing both correctness and efficiency has become crucial. Current methods primarily focus on correctness, often overlooking efficiency. To address this gap, we introduce EffiCoder to improve both aspects by fine-tuning LLMs on a high-quality dataset comprising correct and efficient code samples. Our methodology involves leveraging multiple LLMs to generate diverse candidate code solutions for various tasks across different programming languages. We then evaluate these solutions by measuring their execution time and memory usage through local execution. The code solution with the lowest execution time and memory consumption is selected as the final output for each task. Experimental results demonstrate significant improvements when fine-tuning with Effi-Instruct. For instance, Qwen2.5-Coder-7B-Instruct's pass@1 score increases from 44.8\% to 57.7\%, while the average execution time for correct tasks decreases by 48.4\%. EffiCoder offers a scalable and effective solution for advancing AI-driven code generation, benefiting software development and computational problem-solving. The source code of Effi-Code was released at https://github.com/huangd1999/EffiCoder.

Mingzhe Du, Luu Anh Tuan, Yue Liu et al.

Large Language Models (LLMs) generate functionally correct solutions but often fall short in code efficiency, a critical bottleneck for real-world deployment. In this paper, we introduce a novel test-time iterative optimization framework to address this, employing a closed-loop system where LLMs iteratively refine code based on empirical performance feedback from an execution sandbox. We explore three training strategies: Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Group Relative Policy Optimization (GRPO). Experiments on our Venus dataset and the APPS benchmark show that SFT and DPO rapidly saturate in efficiency gains. In contrast, GRPO, using reinforcement learning (RL) with execution feedback, continuously optimizes code performance, significantly boosting both pass@1 (from 47% to 62%) and the likelihood of outperforming human submissions in efficiency (from 31% to 45%). Our work demonstrates effective test-time code efficiency improvement and critically reveals the power of RL in teaching LLMs to truly self-improve code efficiency.

Dong Huang, Jie M. Zhang, Michael Luck et al.

The advancement of natural language processing (NLP) has been significantly boosted by the development of transformer-based large language models (LLMs). These models have revolutionized NLP tasks, particularly in code generation, aiding developers in creating software with enhanced efficiency. Despite their advancements, challenges in balancing code snippet generation with effective test case generation and execution persist. To address these issues, this paper introduces Multi-Agent Assistant Code Generation (AgentCoder), a novel solution comprising a multi-agent framework with specialized agents: the programmer agent, the test designer agent, and the test executor agent. During the coding procedure, the programmer agent will focus on the code generation and refinement based on the test executor agent's feedback. The test designer agent will generate test cases for the generated code, and the test executor agent will run the code with the test cases and write the feedback to the programmer. This collaborative system ensures robust code generation, surpassing the limitations of single-agent models and traditional methodologies. Our extensive experiments on 9 code generation models and 12 enhancement approaches showcase AgentCoder's superior performance over existing code generation models and prompt engineering techniques across various benchmarks. For example, AgentCoder (GPT-4) achieves 96.3\% and 91.8\% pass@1 in HumanEval and MBPP datasets with an overall token overhead of 56.9K and 66.3K, while state-of-the-art obtains only 90.2\% and 78.9\% pass@1 with an overall token overhead of 138.2K and 206.5K.