Provable Test-Time Scaling for Beam Search in LLM Reasoning
Beam-search-based test-time methods provide an effective way to improve large language model (LLM) performance on long-horizon generation by pruning invalid reasoning paths early, leading to significantly improved reasoning efficiency and more favorable test-time cost scaling. Despite strong empirical success, the theo...
Qi-Jia He, Yu Huang, Yuan Cheng et al.
· 0 citations