CodeT: Code Generation with Generated Tests - OpenReview
Feb 1, 2023 · In this paper, we propose a novel method, CodeT, that leverages the same pre-trained language models to automatically generate test cases for the code samples, thus …
CodeMMLU: A Multi-Task Benchmark for Assessing Code …
Jan 22, 2025 · Abstract: Recent advances in Code Large Language Models (CodeLLMs) have primarily focused on open-ended code generation, often overlooking the crucial aspect of code …
Revisiting Chain-of-Thought in Code Generation: Do Language …
May 1, 2025 · Large Language Models (LLMs) have demonstrated exceptional performance in code generation, becoming increasingly vital for software engineering and development. …
GraphCodeBERT: Pre-training Code Representations with Data …
Jan 12, 2021 · We evaluate our model on four tasks, including code search, clone detection, code translation, and code refinement. Results show that code structure and newly introduced pre …
RedCodeAgent: Automatic Red-teaming Agent against Code Agents
Sep 28, 2024 · By autonomously exploring and exploiting vulnerabilities of code agents, RedCodeAgent provides critical insights into the evolving security risks of code agents.
SemCoder: Training Code Language Models with Comprehensive...
We propose training Code LLMs not only to write code but also to understand code semantics by reasoning about key properties, constraints, and execution behaviors using natural language, …
Code2JSON: Can a Zero-Shot LLM Extract Code Features for Code …
Mar 6, 2025 · To address this, we introduce CODE2JSON, a zero-shot technique that leverages LLMs for extracting NL representations from code via semantic parsing. CODE2JSON serves …
LeDex: Training LLMs to Better Self-Debug and Explain Code
Sep 25, 2024 · In this work, we propose LeDex, a training framework that significantly improves the self-debugging capability of LLMs. Intuitively, we observe that a chain of explanations on …
CoDe: Blockwise Control for Denoising Diffusion Models
Apr 21, 2025 · In this work, we explore a simple inference-time gradient-free guidance approach, called controlled denoising (CoDe), that circumvents the need for differentiable guidance …
Self-play with Execution Feedback: Improving Instruction-following ...
Jan 22, 2025 · The main novelty of this work is that it reduces instruction-following alignment to a verifiable process using code. AUTOIF leverages execution feedback from code verification …