c42c8d51556fabb4b57fc86d3d3d0d09-Paper-Datasets_and_Benchmarks_Track.pdf

QuestBench: acquire inf Can ormation LLMs ask in reasoning the right tasks? question to Lar ingly ge being language applied models to reasoning (LLMs) tasks are increassuch as math ning/coding tions typically [15, 34 [ , 18 46 assume , ], 59 logic , 63 all , 6 [ necessary 70 , 10 , 12 ]. Users orld scenarios may omit often crucial violate details this in in such en math cas vironme es, problems, LLMs nts with need and partial the robots ability observ might to proacti ability operate v .