Where Animacy Lives in Large Language Models: Tracing the Circuits of the Animacy Concept

2026-07-23

AI Summary

This study investigates how large language models distinguish between animate and inanimate verbs and examines whether they rely on locally interpretable causal mechanisms for such judgments. By constructing a minimal-pair dataset and applying circuit discovery, deep ablation, and cross-model comparative analysis across four open-source models, the work identifies—for the first time—a causal circuit responsible for animacy classification. The findings reveal that this circuit is distributed, context-dependent, and only partially generalizes across contexts, indicating that animacy representations are non-discrete and highly sensitive to linguistic context. These results challenge prevailing assumptions of highly localized conceptual circuits in neural language models.

Abstract

Distinguishing animate from inanimate concepts in written language requires more than shallow text processing, as it involves recognizing complex selectional constraints and contextual cues, such as verb-argument interactions. Yet, current large language models (LLMs) appear to be capable of doing it. We investigate whether this animacy-sensitive behavior of LLMs can be traced to a localized set of causally relevant components and connections. To do so, we construct a controlled dataset of minimal pairs and perform circuit discovery on four open-weight models. Through in-depth experiments and ablations, we show that a causal mechanism responsible for handling animacy in these models does exist, thus discovering an animacy circuit. At the same time, this circuit appears to be less localized compared to other known ones and generalizes only partially across models and animacy tasks, confirming the distributed, context-dependent, and somewhat graded nature of the animacy concept.

Problem

• animacy
• large language models
• concept representation
• circuit discovery
• selectional constraints

Innovation

• animacy circuit
• circuit discovery
• large language models
• causal mechanism
• distributed representation