Beyond the Imitation Game Benchmark
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to probe large language models and extrapolate their future capabilities. Big-bench include more than 200 tasks.
Image source: https://arxiv.org/pdf/2206.04615.pdf
Variants: BBH-nlp, BBH-alg, Big-bench Hard, BIG-bench (Logical Sequence), BIG-bench (Logical Fallacy Detection), BIG-bench (Known Unknowns), BIG-bench (Hindu Knowledge), BIG-bench (Novel Concepts), BIG-bench (StrategyQA), BIG-bench (Winowhy), BIG-bench (Logic Grid Puzzle), BIG-bench (Anachronisms), BIG-bench (Temporal Sequences), BIG-bench (Sports Understanding), BIG-bench (SNARKS), BIG-bench (Ruin Names), BIG-bench (Reasoning About Colored Objects), BIG-bench (Penguins In A Table), BIG-bench (Navigate), BIG-bench (Movie Recommendation), BIG-bench (Hyperbaton), BIG-bench (Formal Fallacies Syllogisms Negation), BIG-bench (Disambiguation QA), BIG-bench (Date Understanding), BIG-bench (Causal Judgment), BIG-bench-lite, Big-bench Lite, BIG-bench
This dataset is used in 41 benchmarks:
Recent papers with results on this dataset: