BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent

TL;DR

BrowseComp-Plus introduces a controlled, fair, and transparent benchmark for evaluating deep research agents by using a fixed corpus and human-verified documents, enabling better comparison and analysis of system components.

Contribution

It presents BrowseComp-Plus, a new benchmark with curated data for fair, reproducible evaluation of deep research agents, addressing limitations of existing web API-based benchmarks.

Findings

BrowseComp-Plus effectively distinguishes deep research system performance.

Open-source Search-R1 with BM25 achieves 3.86% accuracy.

GPT-5 with Qwen3-Embedding-8B retriever reaches 70.1% accuracy.

Abstract

Deep-Research agents, which integrate large language models (LLMs) with search tools, have shown success in improving the effectiveness of handling complex queries that require iterative search planning and reasoning over search results. Evaluations on current benchmarks like BrowseComp relies on black-box live web search APIs, have notable limitations in (1) fairness: dynamic and opaque web APIs hinder fair comparisons and reproducibility of deep research methods; (2) transparency: lack of control over the document corpus makes it difficult to isolate retriever contributions. In other words, the current evaluations may compare a complete deep research system at a given time, but they do not foster well-controlled experiments to provide insights into the capability of underlying deep research LLMs. To address these challenges, we introduce BrowseComp-Plus, a benchmark derived from BrowseComp, employing a fixed, carefully curated corpus. Each query in BrowseComp-Plus includes human-verified supporting documents and mined challenging negatives, enabling controlled experimentation. The benchmark is shown to be effective in distinguishing the performance of deep research systems. For instance, the open-source model Search-R1, when paired with the BM25 retriever, achieves 3.86% accuracy, whereas the GPT-5 achieves 55.9%. Integrating the GPT-5 with the Qwen3-Embedding-8B retriever further enhances its accuracy to 70.1% with fewer search calls. This benchmark allows comprehensive evaluation and disentangled analysis of deep research agents and retrieval methods, fostering insights into retrieval effectiveness, citation accuracy, and context engineering in Deep-Research system.