Benchmark Almanac

Baidu

We track 4 benchmarks that Baidu made or helped make. Its models are on its lab page.

2 in use 3 with public code 1 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
MarathonDocBench Models parse 100 documents spanning 103 to 300 pages. Evaluation scores content, cross-page continuations, figures, reading order, and heading hierarchy. Documents, charts and text in images Oct 2026 77.1% In use
Real5-OmniDocBench Models parse the same 1,355 document pages under five physical capture conditions. Text edit distance, formula matching, and table similarity determine the overall score. Documents, charts and text in images Mar 2026 93.7% In use
DuMateBench Agents complete real-world productivity and office-work tasks in sandboxed workspaces with tools. Task evaluators check the requested artifact and record a reward. Running a business Aug 2026 Not in cards
ToolRet Models retrieve relevant tools for 7,600 requests from a corpus of about 43,000 tools. Evaluation measures ranking quality, precision, recall, and complete tool-set retrieval. Function calling Mar 2025 Not in cards