Preprint
Aug 2026
Thought-Level Beam Search for Reasoning
By periodically pruning unpromising trajectories and immediately branching from high-quality prefixes, Gambit dynamically concentrates compute onto the most promising reasoning traces via a light-weight scorer probing hidden states while maintaining continuous high hardware utilization.
Lijie Yang, Hongyin Luo, Tri Dao et al.
· 0 citations