作者位次 #6/13ACL 2026Poster· Natural Language Processing
T ool PRM : Fine-Grained Inference Scaling of Structured Outputs for Function Calling
Ruiming Tang (Huawei Noah' · Yuanyuan Shi · Xin Peng · Renjie Ding · Hairui Wang · Yuxuan Peng · Bizhe Bai · Weixi Song · Fengshuo Bai · Huacan Chai · Ruiming Tang (Huawei Noah' · Fei Huang · Ying Wen
摘要
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 ToolPRM, a process reward model scoring each intra-call...