Toward a Gricean Retreat: Probing LLMs for Knowledge Boundaries and Referent Specificity

2026-08-13

AI Summary

This work investigates how large language models respond when queried about entities beyond their knowledge boundaries, revealing a tendency to generate fabricated specifics rather than defaulting to safer, generalized statements. The study introduces, for the first time in model alignment research, a specificity fallback mechanism inspired by Grice’s Cooperative Principle. A novel evaluation benchmark built upon T-REx is developed to probe models’ awareness of their knowledge limits and their generation strategies by systematically manipulating entity familiarity and referential specificity. Experimental results demonstrate that while models can accurately recognize knowledge boundaries and predict appropriate levels of output specificity, they consistently favor concrete references during generation and rarely select correct generalized alternatives, exposing a critical misalignment between their internal knowledge representations and actual generation behavior.

Abstract

When asked about entities outside their knowledge boundary, LLMs routinely fabricate plausible-sounding details rather than backing off to safer, more general claims. We frame this failure through a Gricean lens: a cooperative speaker who is uncertain about a referent retreats up the specificity hierarchy, trading informativeness for truthfulness. We ask whether LLMs have the ingredients to perform this retreat. Using a T-REx-based benchmark that varies entity familiarity and referent specificity, we probe models to answer two questions: (i) do their activations encode whether a referent falls inside the knowledge boundary, and (ii) do they anticipate the specificity of the referent they are about to generate? We find that the answer to both is yes, but the two signals are not reconciled in generation. Models overwhelmingly prefer specific referents even when the entity is unknown to them, and do so even when offered correct generic alternatives. The substrate for a Gricean retreat is present, but the policy that would act on it is not. We position our findings as a first step toward Gricean alignment, training or steering objectives that couple knowledge-boundary awareness to referent-specificity during generation.

Problem

• knowledge boundaries
• referent specificity
• Gricean retreat
• large language models
• hallucination

Innovation

• Gricean retreat
• knowledge boundary
• referent specificity
• hallucination mitigation
• language model alignment