一作ACL 2026Poster· Natural Language Processing
PRIME : A Process-Outcome Alignment Benchmark for Verifiable Reasoning in Mathematics and Engineering
Xiangfeng Wang · Hangyu Guo · Yanlin Lai · Mitt Huang · Liang Zhao · Chengyuan Yao · Yinmin Zhang · Qi Han · Xiaoxiaoren · Chun Yuan · an Jiaotong University) · Zheng Ge · Xiangyu Zhang · Daxin Jiang
摘要
While model-based verifiers are essential for scaling Reinforcement Learning with Verifiable Rewards (RLVR), current outcome-centric verification paradigms primarily focus on the consistency between the final result and the ground truth, often neglecting potential errors in the derivation process...