What it measures

6,216 fine-grained step-level error problems evaluating PRMs across simplicity, soundness, and sensitivity (non-redundancy, circular logic, phantom steps).

Reading the measurement

F1 balances precision and recall. Check both: a system can miss relevant evidence or include unsupported material, and those errors have different consequences. Keep the matching rule, averaging policy and denominator fixed.

Published evaluation

Published PRMScore evaluates process-level reward judgments across simplicity, soundness and sensitivity. Critic models and reward models retain their source labels.

Where this fits in an evaluation suite

Format compliance, preference, truthfulness, and safety require separate measurements. Keep desirable writing style from masking factual errors.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Published PRMScore evaluates process-level reward judgments across simplicity, soundness and sensitivity. Critic models and reward models retain their source labels.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. PRMBench: A Fine-grained and Challenging Benchmark for Process-Level Reward Models — published result table ↗

    Song, Mingyang, Su, Zhaochen, Qu, Xiaoye, Zhou, Jiawei, Cheng, Yu · 2025

Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.