TY  - RPRT
TI  - CRAB-Bench: Evaluating LLM Agents under Complex Task Dependencies and Human-aligned User Simulation
AU  - Danqing Wang
AU  - Akshay Sivaraman
AU  - Lei Li
PY  - 2026
UR  - https://arxiv.org/abs/2606.01815
ID  - 2606.01815
ER  -