BehaviorBench: New benchmark for modeling user decisions from behavioral traces

Arxiv CS.AI (top papers)

BehaviorBench evaluates AI models on predicting real-world user decisions from behavioral traces. The benchmark uses actual user interaction data rather than simulated behavior.

benchmarks

Arxiv CS.AI (top papers)·Liangwei YangJielin QiuZixiang ChenMing ZhuJuntao TanZhiwei LiuWenting ZhaoZhujun LanAkshara PrabhakarSilvio SavareseHuan WangShelby Heinecke

8 days ago