TY  - RPRT
TI  - How Many Tasks Are Enough for Agent Benchmark Decisions? A Replay Analysis of Public LLM Agent Benchmarks
AU  - Wei-Jung Huang
PY  - 2026
UR  - https://arxiv.org/abs/2607.12338
ID  - 2607.12338
ER  -