Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Tree Search

3 points| vrm | 8 months ago |arxiv.org