1<?xml version='1.0' encoding='UTF-8'?>2<feed xmlns:opensearch="http://a9.com/-/spec/opensearch/1.1/" xmlns:arxiv="http://arxiv.org/schemas/atom" xmlns="http://www.w3.org/2005/Atom">3 <id>https://arxiv.org/api/ttPaYvImU7mSRia0id58PCFWnIg</id>4 <title>arXiv Query: search_query=&id_list=2504.08066,2501.04227,2410.05080&start=0&max_results=3</title>5 <updated>2026-10-30T03:04:25Z</updated>6 <link href="https://arxiv.org/api/query?search_query=&start=0&max_results=3&id_list=2504.08066,2501.04227,2410.05080" type="application/atom+xml"/>7 <opensearch:itemsPerPage>3</opensearch:itemsPerPage>8 <opensearch:totalResults>3</opensearch:totalResults>9 <opensearch:startIndex>0</opensearch:startIndex>10 <entry>11 <id>http://arxiv.org/abs/2504.08066v1</id>12 <title>The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search</title>13 <updated>2025-04-10T18:44:41Z</updated>14 <link href="https://arxiv.org/abs/2504.08066v1" rel="alternate" type="text/html"/>15 <link href="https://arxiv.org/pdf/2504.08066v1" rel="related" type="application/pdf" title="pdf"/>16 <summary>AI is increasingly playing a pivotal role in transforming how scientific discoveries are made. We introduce The AI Scientist-v2, an end-to-end agentic system capable of producing the first entirely AI generated peer-review-accepted workshop paper. This system iteratively formulates scientific hypotheses, designs and executes experiments, analyzes and visualizes data, and autonomously authors scientific manuscripts. Compared to its predecessor (v1, Lu et al., 2024 arXiv:2408.06292), The AI Scientist-v2 eliminates the reliance on human-authored code templates, generalizes effectively across diverse machine learning domains, and leverages a novel progressive agentic tree-search methodology managed by a dedicated experiment manager agent. Additionally, we enhance the AI reviewer component by integrating a Vision-Language Model (VLM) feedback loop for iterative refinement of content and aesthetics of the figures. We evaluated The AI Scientist-v2 by submitting three fully autonomous manuscripts to a peer-reviewed ICLR workshop. Notably, one manuscript achieved high enough scores to exceed the average human acceptance threshold, marking the first instance of a fully AI-generated paper successfully navigating a peer review. This accomplishment highlights the growing capability of AI in conducting all aspects of scientific research. We anticipate that further advancements in autonomous scientific discovery technologies will profoundly impact human knowledge generation, enabling unprecedented scalability in research productivity and significantly accelerating scientific breakthroughs, greatly benefiting society at large. We have open-sourced the code at https://github.com/SakanaAI/AI-Scientist-v2 to foster the future development of this transformative technology. We also discuss the role of AI in science, including AI safety.</summary>17 <category term="cs.AI" scheme="http://arxiv.org/schemas/atom"/>18 <category term="cs.CL" scheme="http://arxiv.org/schemas/atom"/>19 <category term="cs.LG" scheme="http://arxiv.org/schemas/atom"/>20 <published>2025-04-10T18:44:41Z</published>21 <arxiv:primary_category term="cs.AI"/>22 <author>23 <name>Yutaro Yamada</name>24 </author>25 <author>26 <name>Robert Tjarko Lange</name>27 </author>28 <author>29 <name>Cong Lu</name>30 </author>31 <author>32 <name>Shengran Hu</name>33 </author>34 <author>35 <name>Chris Lu</name>36 </author>37 <author>38 <name>Jakob Foerster</name>39 </author>40 <author>41 <name>Jeff Clune</name>42 </author>43 <author>44 <name>David Ha</name>45 </author>46 </entry>47 <entry>48 <id>http://arxiv.org/abs/2501.04227v2</id>49 <title>Agent Laboratory: Using LLM Agents as Research Assistants</title>50 <updated>2025-06-17T16:19:14Z</updated>51 <link href="https://arxiv.org/abs/2501.04227v2" rel="alternate" type="text/html"/>52 <link href="https://arxiv.org/pdf/2501.04227v2" rel="related" type="application/pdf" title="pdf"/>53 <summary>Historically, scientific discovery has been a lengthy and costly process, demanding substantial time and resources from initial conception to final results. To accelerate scientific discovery, reduce research costs, and improve research quality, we introduce Agent Laboratory, an autonomous LLM-based framework capable of completing the entire research process. This framework accepts a human-provided research idea and progresses through three stages--literature review, experimentation, and report writing to produce comprehensive research outputs, including a code repository and a research report, while enabling users to provide feedback and guidance at each stage. We deploy Agent Laboratory with various state-of-the-art LLMs and invite multiple researchers to assess its quality by participating in a survey, providing human feedback to guide the research process, and then evaluate the final paper. We found that: (1) Agent Laboratory driven by o1-preview generates the best research outcomes; (2) The generated machine learning code is able to achieve state-of-the-art performance compared to existing methods; (3) Human involvement, providing feedback at each stage, significantly improves the overall quality of research; (4) Agent Laboratory significantly reduces research expenses, achieving an 84% decrease compared to previous autonomous research methods. We hope Agent Laboratory enables researchers to allocate more effort toward creative ideation rather than low-level coding and writing, ultimately accelerating scientific discovery.</summary>54 <category term="cs.HC" scheme="http://arxiv.org/schemas/atom"/>55 <category term="cs.AI" scheme="http://arxiv.org/schemas/atom"/>56 <category term="cs.CL" scheme="http://arxiv.org/schemas/atom"/>57 <category term="cs.LG" scheme="http://arxiv.org/schemas/atom"/>58 <published>2025-01-08T01:58:42Z</published>59 <arxiv:primary_category term="cs.HC"/>60 <author>61 <name>Samuel Schmidgall</name>62 </author>63 <author>64 <name>Yusheng Su</name>65 </author>66 <author>67 <name>Ze Wang</name>68 </author>69 <author>70 <name>Ximeng Sun</name>71 </author>72 <author>73 <name>Jialian Wu</name>74 </author>75 <author>76 <name>Xiaodong Yu</name>77 </author>78 <author>79 <name>Jiang Liu</name>80 </author>81 <author>82 <name>Michael Moor</name>83 </author>84 <author>85 <name>Zicheng Liu</name>86 </author>87 <author>88 <name>Emad Barsoum</name>89 </author>90 </entry>91 <entry>92 <id>http://arxiv.org/abs/2410.05080v3</id>93 <title>ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery</title>94 <updated>2025-03-31T14:39:44Z</updated>95 <link href="https://arxiv.org/abs/2410.05080v3" rel="alternate" type="text/html"/>96 <link href="https://arxiv.org/pdf/2410.05080v3" rel="related" type="application/pdf" title="pdf"/>97 <summary>The advancements of large language models (LLMs) have piqued growing interest in developing LLM-based language agents to automate scientific discovery end-to-end, which has sparked both excitement and skepticism about their true capabilities. In this work, we call for rigorous assessment of agents on individual tasks in a scientific workflow before making bold claims on end-to-end automation. To this end, we present ScienceAgentBench, a new benchmark for evaluating language agents for data-driven scientific discovery. To ensure the scientific authenticity and real-world relevance of our benchmark, we extract 102 tasks from 44 peer-reviewed publications in four disciplines and engage nine subject matter experts to validate them. We unify the target output for every task to a self-contained Python program file and employ an array of evaluation metrics to examine the generated programs, execution results, and costs. Each task goes through multiple rounds of manual validation by annotators and subject matter experts to ensure its annotation quality and scientific plausibility. We also propose two effective strategies to mitigate data contamination concerns. Using ScienceAgentBench, we evaluate five open-weight and proprietary LLMs, each with three frameworks: direct prompting, OpenHands CodeAct, and self-debug. Given three attempts for each task, the best-performing agent can only solve 32.4% of the tasks independently and 34.3% with expert-provided knowledge. In addition, we evaluate OpenAI o1-preview with direct prompting and self-debug, which can boost the performance to 42.2%, demonstrating the effectiveness of increasing inference-time compute but with more than 10 times the cost of other LLMs. Still, our results underscore the limitations of current language agents in generating code for data-driven discovery, let alone end-to-end automation for scientific research.</summary>98 <category term="cs.CL" scheme="http://arxiv.org/schemas/atom"/>99 <category term="cs.AI" scheme="http://arxiv.org/schemas/atom"/>100 <category term="cs.LG" scheme="http://arxiv.org/schemas/atom"/>101 <published>2024-10-07T14:33:50Z</published>102 <arxiv:comment>ICLR 2025. 60 pages</arxiv:comment>103 <arxiv:primary_category term="cs.CL"/>104 <author>105 <name>Ziru Chen</name>106 </author>107 <author>108 <name>Shijie Chen</name>109 </author>110 <author>111 <name>Yuting Ning</name>112 </author>113 <author>114 <name>Qianheng Zhang</name>115 </author>116 <author>117 <name>Boshi Wang</name>118 </author>119 <author>120 <name>Botao Yu</name>121 </author>122 <author>123 <name>Yifei Li</name>124 </author>125 <author>126 <name>Zeyi Liao</name>127 </author>128 <author>129 <name>Chen Wei</name>130 </author>131 <author>132 <name>Zitong Lu</name>133 </author>134 <author>135 <name>Vishal Dey</name>136 </author>137 <author>138 <name>Mingyi Xue</name>139 </author>140 <author>141 <name>Frazier N. Baker</name>142 </author>143 <author>144 <name>Benjamin Burns</name>145 </author>146 <author>147 <name>Daniel Adu-Ampratwum</name>148 </author>149 <author>150 <name>Xuhui Huang</name>151 </author>152 <author>153 <name>Xia Ning</name>154 </author>155 <author>156 <name>Song Gao</name>157 </author>158 <author>159 <name>Yu Su</name>160 </author>161 <author>162 <name>Huan Sun</name>163 </author>164 </entry>165</feed>166