LlamaIndex (@llama_index) and Kaggle launched ExtractBench, a schema-guided document-extraction leaderboard for downstream agents. It covers the documents most likely to break workflows—long record lists, noisy scans, handwriting, and complex tables—across 370 enterprise documents, 8 business domains, and 67 document types, comparing LlamaParse, Codex, Claude Code, and more. The leaderboard is live on Kaggle.

Key Takeaways

  • ExtractBench is live on Kaggle for schema-guided enterprise document extraction.
  • 370 documents across 8 domains and 67 types, including scans, handwriting, long lists, and complex tables.
  • Compares LlamaParse, Codex, Claude Code, and more for downstream agent workflows.
ADSponsored