Unverified50% confidenceReleaseExact time
The researchers built the UtilityBench benchmark, containing 15,000 instances with verifiable answers.
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers constructed TrialPD-Bench, a benchmark set annotated by experts, containing 150 public clinical trial protocol PDFs and covering four types of tasks.69% similarUnverifiedThe test ran a total of 130 variants, covering different levels of reasoning effort, called via OpenRouter with official endpoints for each lab locked in wherever possible.64% similarUnverifiedThe researchers used 15 runs to evaluate the frequency with which the bound holds, rather than relying on a single experiment.63% similarUnverifiedThe experiment tested 15 frontier models.62% similarUnverifiedThe author reviewed approximately 1,500 detection calls, manually mined false positives, adjusted bounding boxes, and then retrained the model.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073847API
curl https://kongchang.com/api/v1/knowledge/claims/1073847MCP
get_claim(id=1073847)