PaperBench: Evaluating AI's Ability to Replicate AI Research

• URL: http://arxiv.org/abs/2504.01848v1
• Date: Wed, 02 Apr 2025 15:55:24 GMT
• Title: PaperBench: Evaluating AI's Ability to Replicate AI Research
• Authors: Giulio Starace, Oliver Jaffe, Dane Sherburn, James Aung, Jun Shern Chan, Leon Maksin, Rachel Dias, Evan Mays, Benjamin Kinsella, Wyatt Thompson, Johannes Heidecke, Amelia Glaese, Tejal Patwardhan,
• Abstract summary: 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.
• Score: 3.4567792239799133
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We introduce PaperBench, a benchmark evaluating the ability of AI agents to replicate state-of-the-art AI research. Agents must replicate 20 ICML 2024 Spotlight and Oral papers from scratch, including understanding paper contributions, developing a codebase, and successfully executing experiments. For objective evaluation, we develop rubrics that hierarchically decompose each replication task into smaller sub-tasks with clear grading criteria. In total, PaperBench contains 8,316 individually gradable tasks. Rubrics are co-developed with the author(s) of each ICML paper for accuracy and realism. To enable scalable evaluation, we also develop an LLM-based judge to automatically grade replication attempts against rubrics, and assess our judge's performance by creating a separate benchmark for judges. We evaluate several frontier models on PaperBench, finding that the best-performing tested agent, Claude 3.5 Sonnet (New) with open-source scaffolding, achieves an average replication score of 21.0\%. Finally, we recruit top ML PhDs to attempt a subset of PaperBench, finding that models do not yet outperform the human baseline. We \href{https://github.com/openai/preparedness}{open-source our code} to facilitate future research in understanding the AI engineering capabilities of AI agents.

Related papers

• Can LLMs Generate Tabular Summaries of Science Papers? Rethinking the Evaluation Protocol [83.90769864167301]
  Literature review tables are essential for summarizing and comparing collections of scientific papers. We explore the task of generating tables that best fulfill a user's informational needs given a collection of scientific papers. Our contributions focus on three key challenges encountered in real-world use: (i) User prompts are often under-specified; (ii) Retrieved candidate papers frequently contain irrelevant content; and (iii) Task evaluation should move beyond shallow text similarity techniques.
  arXiv  Detail & Related papers  (2025-04-14T14:52:28Z)
• AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories [59.214178488091584]
  We propose AgentRewardBench, the first benchmark to assess the effectiveness of LLM judges for evaluating web agents. Using our benchmark, we evaluate 12 LLM judges and find that no single LLM excels across all benchmarks. We also find that the rule-based evaluation used by common benchmarks tends to underreport the success rate of web agents.
  arXiv  Detail & Related papers  (2025-04-11T19:49:22Z)
• SciReplicate-Bench: Benchmarking LLMs in Agent-driven Algorithmic Reproduction from Research Papers [16.80818230868491]
  This study evaluates large language models (LLMs) in generating code from algorithm descriptions from recent NLP papers.<n>To facilitate rigorous evaluation, we introduce SciReplicate-Bench, a benchmark of 100 tasks from 36 NLP papers published in 2024.<n>Building on SciReplicate-Bench, we propose Sci-Reproducer, a multi-agent framework consisting of a Paper Agent that interprets algorithmic concepts from literature and a Code Agent that retrieves dependencies from repositories and implement solutions.
  arXiv  Detail & Related papers  (2025-03-31T22:02:24Z)
• 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)
• CycleResearcher: Improving Automated Research via Automated Review [37.03497673861402]
  This paper explores the possibility of using open-source post-trained large language models (LLMs) as autonomous agents capable of performing the full cycle of automated research and review. To train these models, we develop two new datasets, reflecting real-world machine learning research and peer review dynamics. Our results demonstrate that CycleReviewer achieves promising performance with a 26.89% reduction in mean absolute error (MAE) compared to individual human reviewers in predicting paper scores.
  arXiv  Detail & Related papers  (2024-10-28T08:10:21Z)
• AutoPenBench: Benchmarking Generative Agents for Penetration Testing [42.681170697805726]
  This paper introduces AutoPenBench, an open benchmark for evaluating generative agents in automated penetration testing. We present a comprehensive framework that includes 33 tasks, each representing a vulnerable system that the agent has to attack. We show the benefits of AutoPenBench by testing two agent architectures: a fully autonomous and a semi-autonomous supporting human interaction.
  arXiv  Detail & Related papers  (2024-10-04T08:24:15Z)
• Are Large Language Models Good Classifiers? A Study on Edit Intent Classification in Scientific Document Revisions [62.12545440385489]
  Large language models (LLMs) have brought substantial advancements in text generation, but their potential for enhancing classification tasks remains underexplored. We propose a framework for thoroughly investigating fine-tuning LLMs for classification, including both generation- and encoding-based approaches. We instantiate this framework in edit intent classification (EIC), a challenging and underexplored classification task.
  arXiv  Detail & Related papers  (2024-10-02T20:48:28Z)
• 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)
• UMBRELA: UMbrela is the (Open-Source Reproduction of the) Bing RELevance Assessor [51.20527342770299]
  UMBRELA is an open-source toolkit that reproduces the results of Thomas et al. using OpenAI's GPT-4o model. Our toolkit is designed to be easily studying and can be integrated into existing multi-stage retrieval and evaluation pipelines. UMBRELA will be used in the TREC 2024 RAG Track to aid in relevance assessments.
  arXiv  Detail & Related papers  (2024-06-10T17:58:29Z)
• Skill-Mix: a Flexible and Expandable Family of Evaluations for AI models [50.11814354654953]
  Key ability of an AI agent is to flexibly combine, as needed, the basic skills it has learned. This work introduces Skill-Mix, a new evaluation to measure ability to combine skills.
  arXiv  Detail & Related papers  (2023-10-26T16:55:05Z)
• Generative Judge for Evaluating Alignment [84.09815387884753]
  We propose a generative judge with 13B parameters, Auto-J, designed to address these challenges. Our model is trained on user queries and LLM-generated responses under massive real-world scenarios. Experimentally, Auto-J outperforms a series of strong competitors, including both open-source and closed-source models.
  arXiv  Detail & Related papers  (2023-10-09T07:27:15Z)
• 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)
• Generation of Highlights from Research Papers Using Pointer-Generator Networks and SciBERT Embeddings [5.095525589147811]
  We use a pointer-generator network with coverage mechanism and a contextual embedding layer at the input that encodes the input tokens into SciBERT embeddings. We test our model on a benchmark dataset, CSPubSum, and also present MixSub, a new multi-disciplinary corpus of papers for automatic research highlight generation.
  arXiv  Detail & Related papers  (2023-02-14T12:45:14Z)
• AIBench Training: Balanced Industry-Standard AI Training Benchmarking [26.820244556465333]
  Earlier-stage evaluations of a new AI architecture/system need affordable benchmarks. We use real-world benchmarks to cover the factors space that impacts the learning dynamics. We contribute by far the most comprehensive AI training benchmark suite.
  arXiv  Detail & Related papers  (2020-04-30T11:08:49Z)

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.