WildIFEval

WildIFEval Dataset This dataset was originally introduced in the paper WildIFEval: Instruction Following in the Wild, available on arXiv. Code: github.com/gililior/wild-if-eval-code Model predictions & LLM-as-a-judge scores: gililior/wild-if-eval-predictions Dataset Overview The WildIFEval dataset is designed for evaluating instruction-following capabilities in language models. It provides decompositions of conversations extracted from the LMSYS-Chat-1M… See the full description on the dataset page: huggingface.co/datasets/gililior/wild-if-eval.