CodeMonkeys: Scaling Test-Time Compute for Software Engineering

• URL: http://arxiv.org/abs/2501.14723v2
• Date: Mon, 03 Feb 2025 18:57:05 GMT
• Title: CodeMonkeys: Scaling Test-Time Compute for Software Engineering
• Abstract summary: Scaling test-time compute is a promising axis for improving LLM capabilities.<n>Here, we explore this problem in the context of solving real-world GitHub issues from the SWE-bench dataset.<n>Our system, named CodeMonkeys, allows models to iteratively edit a by jointly generating and running a testing script alongside their draft edit.
• Score: 45.84513832258217
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Scaling test-time compute is a promising axis for improving LLM capabilities. However, test-time compute can be scaled in a variety of ways, and effectively combining different approaches remains an active area of research. Here, we explore this problem in the context of solving real-world GitHub issues from the SWE-bench dataset. Our system, named CodeMonkeys, allows models to iteratively edit a codebase by jointly generating and running a testing script alongside their draft edit. We sample many of these multi-turn trajectories for every issue to generate a collection of candidate edits. This approach lets us scale "serial" test-time compute by increasing the number of iterations per trajectory and "parallel" test-time compute by increasing the number of trajectories per problem. With parallel scaling, we can amortize up-front costs across multiple downstream samples, allowing us to identify relevant codebase context using the simple method of letting an LLM read every file. In order to select between candidate edits, we combine voting using model-generated tests with a final multi-turn trajectory dedicated to selection. Overall, CodeMonkeys resolves 57.4% of issues from SWE-bench Verified using a budget of approximately 2300 USD. Our selection method can also be used to combine candidates from different sources. Selecting over an ensemble of edits from existing top SWE-bench Verified submissions obtains a score of 66.2% and outperforms the best member of the ensemble on its own. We fully release our code and data at https://scalingintelligence.stanford.edu/pubs/codemonkeys.

Related papers

• Scaling Test-Time Compute for Agentic Coding [126.72747643609274]
  We propose a test-time scaling framework for agentic coding based on compact representations of rollout trajectories.<n>Our framework converts each rollout into a structured summary that preserves its salient hypotheses, progress, and failure modes.<n>Our method consistently improves the performance of frontier coding agents across SWE-Bench Verified and Terminal-Bench v2.0.
  arXiv  Detail & Related papers  (2026-04-16T17:39:33Z)
• Parallel Test-Time Scaling with Multi-Sequence Verifiers [23.363130292302483]
  Parallel test-time scaling is a powerful technique for improving large language model performance.<n>It is hindered by two key bottlenecks: accurately selecting the correct solution from the candidate pool, and the high inference latency from generating many full solutions.<n>We introduce the Multi-Sequence Verifier (MSV), the first verifier designed to jointly process all candidate solutions and model their interactions.
  arXiv  Detail & Related papers  (2026-03-03T18:53:56Z)
• Increasing LLM Coding Capabilities through Diverse Synthetic Coding Tasks [41.75017840131367]
  Large language models (LLMs) have shown impressive promise in code generation.<n>We present a scalable synthetic data generation pipeline that produces nearly 800k instruction-reasoning-code-test quadruplets.
  arXiv  Detail & Related papers  (2025-10-27T10:54:25Z)
• Learning to Reason Across Parallel Samples for LLM Reasoning [45.60752271688715]
  Scaling test-time compute brings substantial performance gains for large language models.<n>We propose a new way to leverage such multiple sample set.<n>We train a compact LLM, that takes a sequence of multiple samples and output the final answer.
  arXiv  Detail & Related papers  (2025-06-10T17:42:35Z)
• SwingArena: Competitive Programming Arena for Long-context GitHub Issue Solving [90.32201622392137]
  We present SwingArena, a competitive evaluation framework for Large Language Models (LLMs)<n>Unlike traditional static benchmarks, SwingArena models the collaborative process of software by pairing LLMs as iterations, who generate patches, and reviewers, who create test cases and verify the patches through continuous integration (CI) pipelines.
  arXiv  Detail & Related papers  (2025-05-29T18:28:02Z)
• Satori-SWE: Evolutionary Test-Time Scaling for Sample-Efficient Software Engineering [51.7496756448709]
  Language models (LMs) perform well on coding benchmarks but struggle with real-world software engineering tasks.<n>Existing approaches rely on supervised fine-tuning with high-quality data, which is expensive to curate at scale.<n>We propose Test-Time Scaling (EvoScale), a sample-efficient method that treats generation as an evolutionary process.
  arXiv  Detail & Related papers  (2025-05-29T16:15:36Z)
• Sample, Don't Search: Rethinking Test-Time Alignment for Language Models [55.2480439325792]
  We introduce QAlign, a new test-time alignment approach. As we scale test-time compute, QAlign converges to sampling from the optimal aligned distribution for each individual prompt. By adopting recent advances in Markov chain Monte Carlo for text generation, our method enables better-aligned outputs without modifying the underlying model or even requiring logit access.
  arXiv  Detail & Related papers  (2025-04-04T00:41:40Z)
• SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolution [56.9361004704428]
  Large Language Models (LLMs) have demonstrated remarkable proficiency across a variety of complex tasks. SWE-Fixer is a novel open-source framework designed to effectively and efficiently resolve GitHub issues. We assess our approach on the SWE-Bench Lite and Verified benchmarks, achieving state-of-the-art performance among open-source models.
  arXiv  Detail & Related papers  (2025-01-09T07:54:24Z)
• Steering Large Language Models between Code Execution and Textual Reasoning [22.279107036500083]
  Textual reasoning has inherent limitations in solving tasks with challenges in math, logics, optimization, and searching. The recently released OpenAI GPT Code Interpreter and multi-agent frameworks such as AutoGen have demonstrated remarkable proficiency of integrating code generation and execution. We propose three methods to better steer LLM code/text generation and achieve a notable improvement.
  arXiv  Detail & Related papers  (2024-10-04T15:44:47Z)
• Generating Unseen Code Tests In Infinitum [1.0674604700001968]
  We present a method for creating benchmark variations that generalize across coding tasks and programming languages. We implement one benchmark, called textitauto-regression, for the task of text-to-code generation in Python.
  arXiv  Detail & Related papers  (2024-07-29T08:11:20Z)
• Long Code Arena: a Set of Benchmarks for Long-Context Code Models [75.70507534322336]
  Long Code Arena is a suite of six benchmarks for code processing tasks that require project-wide context. These tasks cover different aspects of code processing: library-based code generation, CI builds repair, project-level code completion, commit message generation, bug localization, and module summarization. For each task, we provide a manually verified dataset for testing, an evaluation suite, and open-source baseline solutions.
  arXiv  Detail & Related papers  (2024-06-17T14:58:29Z)
• Nearest Neighbor Speculative Decoding for LLM Generation and Attribution [87.3259169631789]
  Nearest Speculative Decoding (NEST) is capable of incorporating real-world text spans of arbitrary length into the LM generations and providing attribution to their sources. NEST significantly enhances the generation quality and attribution rate of the base LM across a variety of knowledge-intensive tasks. In addition, NEST substantially improves the generation speed, achieving a 1.8x speedup in inference time when applied to Llama-2-Chat 70B.
  arXiv  Detail & Related papers  (2024-05-29T17:55:03Z)
• VerMCTS: Synthesizing Multi-Step Programs using a Verifier, a Large Language Model, and Tree Search [5.389248707675898]
  Large Language Models (LLMs) can generate useful code, but often the code they generate cannot be trusted to be sound. We present VerMCTS, an approach to begin to resolve this issue by generating verified programs in Dafny and Coq.
  arXiv  Detail & Related papers  (2024-02-13T00:55:14Z)
• Distributed Collapsed Gibbs Sampler for Dirichlet Process Mixture Models in Federated Learning [0.22499166814992444]
  This paper proposes a new distributed Markov Chain Monte Carlo (MCMC) inference method for DPMMs (DisCGS) using sufficient statistics. Our approach uses the collapsed Gibbs sampler and is specifically designed to work on distributed data across independent and heterogeneous machines. For instance, with a dataset of 100K data points, the centralized algorithm requires approximately 12 hours to complete 100 iterations while our approach achieves the same number of iterations in just 3 minutes.
  arXiv  Detail & Related papers  (2023-12-18T13:16:18Z)
• Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs [60.58434523646137]
  A popular approach for improving the correctness of output from large language models (LLMs) is Self-Consistency. We introduce Adaptive-Consistency, a cost-efficient, model-agnostic technique that dynamically adjusts the number of samples per question. Our experiments show that Adaptive-Consistency reduces sample budget by up to 7.9 times with an average accuracy drop of less than 0.1%.
  arXiv  Detail & Related papers  (2023-05-19T17:49:25Z)
• Memory-Based Optimization Methods for Model-Agnostic Meta-Learning and Personalized Federated Learning [56.17603785248675]
  Model-agnostic meta-learning (MAML) has become a popular research area. Existing MAML algorithms rely on the episode' idea by sampling a few tasks and data points to update the meta-model at each iteration. This paper proposes memory-based algorithms for MAML that converge with vanishing error.
  arXiv  Detail & Related papers  (2021-06-09T08:47:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.