RExBench: Can coding agents autonomously implement AI research extensions?

• URL: http://arxiv.org/abs/2506.22598v2
• Date: Thu, 17 Jul 2025 18:45:58 GMT
• Title: RExBench: Can coding agents autonomously implement AI research extensions?
• Authors: Nicholas Edwards, Yukyung Lee, Yujun Audrey Mao, Yulu Qin, Sebastian Schuster, Najoung Kim,
• Abstract summary: Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously.<n>We argue that research extension and its implementation is a critical capability for such systems.<n>We introduce RExBench to support the evaluation of this capability.
• Score: 14.147417159347448
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of 12 realistic research experiment implementation tasks that aim to investigate research hypotheses that have not previously been implemented. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate nine LLM agents implemented using three different frameworks: aider, Claude Code, and OpenHands. We find that all agents evaluated fail to autonomously implement the majority of the extensions. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 40%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.

Related papers

• Cognitive Kernel-Pro: A Framework for Deep Research Agents and Agent Foundation Models Training [67.895981259683]
  General AI Agents are increasingly recognized as foundational frameworks for the next generation of artificial intelligence.<n>Current agent systems are either closed-source or heavily reliant on a variety of paid APIs and proprietary tools.<n>We present Cognitive Kernel-Pro, a fully open-source and (to the maximum extent) free multi-module agent framework.
  arXiv  Detail & Related papers  (2025-08-01T08:11:31Z)
• REPRO-Bench: Can Agentic AI Systems Assess the Reproducibility of Social Science Research? [2.111102681327218]
  Existing benchmarks for reproducing research papers focus solely on reproducing results using provided code and data.<n>We introduce REPRO-Bench, a collection of 112 task instances, each representing a social science paper with a publicly available reproduction report.<n>We evaluate three representative AI agents on REPRO-Bench, with the best-performing agent achieving an accuracy of only 21.4%.
  arXiv  Detail & Related papers  (2025-07-25T02:48:30Z)
• From Reproduction to Replication: Evaluating Research Agents with Progressive Code Masking [48.90371827091671]
  AutoExperiment is a benchmark that evaluates AI agents' ability to implement and run machine learning experiments.<n>We evaluate state-of-the-art agents and find that performance degrades rapidly as $n$ increases.<n>Our findings highlight critical challenges in long-horizon code generation, context retrieval, and autonomous experiment execution.
  arXiv  Detail & Related papers  (2025-06-24T15:39:20Z)
• Coding Agents with Multimodal Browsing are Generalist Problem Solvers [48.938445118630284]
  OpenHands-Versa is a generalist AI agent built with a modest number of general tools.<n>We show how existing state-of-the-art multi-agent systems fail to generalize beyond their target domains.
  arXiv  Detail & Related papers  (2025-06-03T15:50:55Z)
• PaperBench: Evaluating AI's Ability to Replicate AI Research [3.4567792239799133]
  PaperBench is a benchmark evaluating the ability of AI agents to replicate state-of-the-art AI research.<n>Agents must replicate 20 ICML 2024 Spotlight and Oral papers from scratch.<n>PaperBench contains 8,316 individually gradable tasks.
  arXiv  Detail & Related papers  (2025-04-02T15:55:24Z)
• MLGym: A New Framework and Benchmark for Advancing AI Research Agents [51.9387884953294]
  We introduce Meta MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing large language models on AI research tasks.<n>This is the first Gym environment for machine learning (ML) tasks, enabling research on reinforcement learning (RL) algorithms for training such agents.<n>We evaluate a number of frontier large language models (LLMs) on our benchmarks such as Claude-3.5-Sonnet, Llama-3.1 405B, GPT-4o, o1-preview, and Gemini-1.5 Pro.
  arXiv  Detail & Related papers  (2025-02-20T12:28:23Z)
• LLM Agents Making Agent Tools [2.5529148902034637]
  Tool use has turned large language models (LLMs) into powerful agents that can perform complex multi-step tasks.<n>But these tools must be implemented in advance by human developers.<n>We propose ToolMaker, an agentic framework that autonomously transforms papers with code into LLM-compatible tools.
  arXiv  Detail & Related papers  (2025-02-17T11:44:11Z)
• The BrowserGym Ecosystem for Web Agent Research [151.90034093362343]
  BrowserGym ecosystem addresses the growing need for efficient evaluation and benchmarking of web agents.<n>We propose an extended BrowserGym-based ecosystem for web agent research, which unifies existing benchmarks from the literature.<n>We conduct the first large-scale, multi-benchmark web agent experiment and compare the performance of 6 state-of-the-art LLMs across 6 popular web agent benchmarks.
  arXiv  Detail & Related papers  (2024-12-06T23:43:59Z)
• CORE-Bench: Fostering the Credibility of Published Research Through a Computational Reproducibility Agent Benchmark [11.794931453828974]
  CORE-Bench is a benchmark consisting of 270 tasks based on 90 scientific papers across three disciplines (computer science, social science, and medicine) We provide an evaluation system to measure the accuracy of agents in a fast and parallelizable way. The best agent achieved an accuracy of 21% on the hardest task, showing the vast scope for improvement in automating routine scientific tasks.
  arXiv  Detail & Related papers  (2024-09-17T17:13:19Z)
• WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks [85.95607119635102]
  Large language models (LLMs) can mimic human-like intelligence.<n>WorkArena++ is designed to evaluate the planning, problem-solving, logical/arithmetic reasoning, retrieval, and contextual understanding abilities of web agents.
  arXiv  Detail & Related papers  (2024-07-07T07:15:49Z)
• AgentQuest: A Modular Benchmark Framework to Measure Progress and Improve LLM Agents [19.439775106707344]
  AgentQuest is a framework where benchmarks and metrics are modular and easily through well documented and easy-to-use APIs. We offer two new evaluation metrics that can reliably track LLM agent progress while solving a task. We exemplify the utility of the metrics on two use cases wherein we identify common failure points and refine the agent architecture to obtain a significant performance increase.
  arXiv  Detail & Related papers  (2024-04-09T16:01:24Z)
• MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation [96.71370747681078]
  We introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate.
  arXiv  Detail & Related papers  (2023-10-05T04:06:12Z)

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.