CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking
arXiv:2610.02460v1 Announce Type: new Abstract: Recent benchmarks rely on user simulators to evaluate AI agents in multi-turn interaction. While existing simulation techniques demonstrate surface fidelity to human style and behavior, ecologically valid interactive benchmarking also requires alignme…
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-10-05 04:00 · arXiv cs.CL
CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking