Benchmarks

Compare state-of-the-art methods across the tasks and datasets used to measure AI progress.

309,131 indexed

Task	Dataset	Domain	Results	Last Update
Mathematical Problem Solving	LiveMath held-out (test)	nlpagents tool use	59	May 25, 2026
Office Application Question Answering	OfficeQA held-out (test)	nlp	59	May 25, 2026
Spreadsheet Manipulation and Reasoning	SpreadsheetBench held-out (test)	tabular time seriesnlp	59	May 25, 2026
Web Search Question Answering	SearchQA held-out (test)	retrieval search rag	59	May 25, 2026
Zero-shot Accuracy	6-task zero-shot (MMLU, PIQA, ARC-E, ARC-C, Winogrande, OBQA)	nlp