# Awesome Process Reward Models

A comprehensive, continually updated collection of process reward models, models that score each reasoning step rather than only the final answer, for mathematical and other LLM tasks.

Awesome Process Reward Models organizes research on process reward models (PRMs), which supervise or score the intermediate steps of a model's reasoning rather than only its final answer, split into sections for PRMs built for mathematical tasks, PRMs for other tasks, other process-supervised models, multimodal PRMs, and benchmarks for evaluating them. Listed mathematical-task entries include methods for debiasing PRMs against response length, single-pass reference-guided annotation, training PRMs without ground-truth process labels, active-learning-based PRM training, PRMs that think through generative reasoning, entropy-driven uncertainty modeling, retrieval-augmented PRMs, and multilingual and universal PRM training frameworks, each tagged with its publication venue and, where available, linked code, models, and data.

The list also tracks related process-supervised approaches outside pure PRMs and dedicated PRM benchmarks, and accepts community contributions to keep the collection current.

Website: https://github.com/RyanLiu112/Awesome-Process-Reward-Models