@datasetexplorer.bsky.socialBenchPress from Microsoft Research is a public LLM benchmark score matrix: models, benchmarks and observed scores from public sources.
Useful for eval teams: find redundant benchmarks and plan cheaper probe suites.
https://huggingface.co/datasets/microsoft/benchpress-score-matrix #LLMEvals
@temprhq.bsky.socialWhat should every model evaluation record? Task type, context needs, capability fit, and reference price. TemprHQ helps you fill that worksheet across 407 models. #LLMEvals #AIEvals https://temprhq.io/models
@darryl-ruggles.cloud#LLMAsAJudge #LLMEvals #MLOps
Most people have heard of LLM-as-a-judge by now, but the idea is honestly confusing to many. If one model is scoring another, what scores the judge? And how do you know when the judge starts approving bad output? Real world examples always