Saved in:
Bibliographic Details
Main Authors: Lin, Jianghao, Shi, Yuanyuan, Peng, Xin, Ding, Renjie, Wang, Hairui, Peng, Yuxuan, Bai, Bizhe, Song, Weixi, Bai, Fengshuo, Chai, Huacan, Zhang, Weinan, Huang, Fei, Wen, Ying
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.14703
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • Large language models (LLMs) excel at function calling, but inference scaling has been explored mainly for unstructured generation. We propose an inference-scaling framework for structured outputs that combines fine-grained beam search with \textbf{ToolPRM}, a process reward model scoring each intra-call decision (function name and argument filling). We build the first fine-grained intra-call supervision dataset via function masking, rollout collection, and step-level annotation. ToolPRM outperforms outcome and coarse-grained reward models in predictive accuracy and yields consistent test-time gains on multiple function-calling benchmarks. We further show that structured generation follows ``\textbf{explore more but retain less}'', since early JSON errors are unrecoverable.