#natural language process...
May 2026
Ask Now, Use Later: Benchmarking the Proactivity Gap in Long-Lived LLM Agents
ATRBench surfaces this proactivity gap in current agents and offers a diagnostic testbed for closing it, and makes it measurable by fixing each user's preferences as hidden ground truth, so success demands asking, not recall.
Binwu Wu, Guanyu Zou, Bingbing Wang et al.
· arXiv.org · 2 citations