Internal signals improve error detection for language model answers

Auditing the Judge: Internal Signals for LLM Reasoning Assurance

Machine Learning

Summary

Language models try to judge when their answers might be wrong to decide which ones need more checking. The authors studied two ways to find mistakes: one from the model’s own confidence signals and another from a probe trained on the model’s internal activity. They found that the probe can catch more errors with a fixed review budget, but it works best when trained on the same type of data. The paper shows that internal signals can help but don’t always generalize well across different tasks.

What this means in practice

  • For ai system engineers: Use internal model signals with supervised probes to prioritize review of language model outputs under strict error-checking budgets.
  • For automated tutoring developers: Improve answer validation in automated math tutoring systems by selectively rechecking responses based on internal model error signals.

Authors

Mayank Gulati, Gerhard Wunder

Abstract

Language-model systems use self-judgment to decide which answers to release, review, or recompute. Because intervention is costly, an assurance score must rank errors well at the available budget, not merely over the full ranking. We compare the judge's continuous verdict-token score with a supervised correctness probe trained on attention-head activations from the same prompt. Exact identities link fixed-budget error recall to residual risk and show that AUROC averages recall over all budgets. Across three open 7-12B models on GSM8K and MATH, the probe finds 1.3-22.5 additional percentage points of errors at 5% review in all six model-dataset settings (median 7.4 points). At 10% routing, probe-guided allocation improves accuracy over equal-cost random allocation in both GSM8K experiments, but beats judge-guided routing only for the weaker judge. Cross-dataset transfer lowers probe AUROC for every model in both directions relative to target-domain training. Because dataset, weight precision, and generation budget change together, this measures a joint protocol shift rather than a dataset-only effect. Internal signals can improve supervised, in-domain budget allocation, but are not domain-general confidence estimators.