Benchmark Almanac

Virginia Tech

We track 3 benchmarks that Virginia Tech made or helped make.

0 in use 1 with public code 2 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
Seal-0 Models answer 111 fact-seeking questions where web search returns conflicting or unhelpful results. Answers are graded for accuracy, usually with search enabled. Questions that need search Jun 2025 57.4% Fading
VQAv2 Open-ended questions about COCO images, balanced to reduce language priors. Scored by consensus accuracy against 10 human answers. General visual questions Dec 2016 88.4% Fading
LLM-SRBench Models infer scientific equations from observations on 239 problems across four scientific domains. Symbolic accuracy checks whether the discovered equation recovers the underlying relationship. Research science Apr 2025 15.1% Not in cards