2025
Step-level Verifier-guided Hybrid Test-Time Scaling for Large Language Models
EMNLP 2025
Test-Time Scaling (TTS) is a promising approach to progressively elicit the model’s intelligence during inference. Recently, training-based TTS methods, such as continued reinforcement learning (RL), have further surged in popularity, while training-free TTS methods are gradually fading from promine