evaluation · 19 topics

why 99 percent accurate agents fail long horizon tasks

tutormoments

tutormoments

artificial analysis endpoint accuracy index

epoch mirrorcode claude fable 5

how to evaluate a new ai model without starting from scratch

ai2 olmo eval workbench

a shared playbook for trustworthy third party evaluations

ai eval compute bottleneck

2026 04 27 papers 2604 22119

sharding prevents llm oversight failures

holocount holistic visual counting benchmark mllms

what current ai benchmarks leave unmeasured

agentstream self evolving llm agents streaming

do models fake alignment without clear consequences

codifying the judge

eta given delta

arxiv 2501 12948

disagreement among frontier llms on real world fact checks