Exgentic/results

该数据集包含通用人工智能代理在多样真实世界基准测试中的详细评估结果，无需特定领域调优。评估覆盖多个代理（如Claude Code、OpenAI Solo、Smolagent、React等）与模型（如Claude Opus 4.5、Gemini 3 Pro、GPT-5.2等）的组合，基准测试包括AppWorld（模拟智能手机环境中的应用任务完成）、BrowseComp+（网页浏览和复杂信息检索）、SWE-bench（真实GitHub仓库的软件工程问题解决）以及TauBench系列（航空、零售、电信领域的客服代理评估）。数据包括基准得分、会话统计、成本、步骤数等指标，用于支持开放代理排行榜。

Detailed evaluation results for general-purpose AI agents across diverse real-world benchmarks — without domain-specific tuning. The dataset covers combinations of agents (e.g., Claude Code, OpenAI Solo, Smolagent, React) and models (e.g., Claude Opus 4.5, Gemini 3 Pro, GPT-5.2), with benchmarks including AppWorld (app-based task completion in simulated smartphone environments), BrowseComp+ (web browsing and complex information retrieval), SWE-bench (software engineering issue resolution on real GitHub repos), and TauBench series (customer service agent evaluation in airline, retail, and telecom domains). Metrics include benchmark scores, session counts, costs, step counts, and more, supporting the Open Agent Leaderboard.