Unverified50% confidenceReleaseExact time
ExtractBench Pro is a newly released benchmark designed specifically for complex "chart + document" reasoning tasks in real-world documents.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
UnverifiedExtractBench Pro incorporates cost into its benchmark evaluation system, promoting evaluation standards closer to production environments.67% similarUnverifiedThe Winograd Schema Challenge was designed to test commonsense reasoning but was later found to contain statistical shortcuts, allowing models to achieve high scores without truly understanding semantics.62% similarUnverifiedτ2-bench is a benchmark specifically designed to evaluate large models' tool-calling capabilities, simulating complex multi-step tool-calling scenarios that agents need to handle in the real world.62% similarUnverifiedCXBench is a project that uses the game of Go to evaluate the capabilities of large language models.61% similarUnverifiedONGRID is a GraphRAG tool that, in addition to retrieval and generation, adds a step for judging whether an answer is trustworthy.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073159API
curl https://kongchang.com/api/v1/knowledge/claims/1073159MCP
get_claim(id=1073159)