Is Inference Mediated by Distinct Semantic Structures in LLMs? A Mechanistic Interpretation
Predicting a label correctly does not necessarily require representing the operation that produces it. Transformer representations are known to carry label-level information, but whether they encode semantic operations producing those labels is unclear. We investigate this in Natural Language Inference using controlled premise-hypothesis pairs that differ by a single semantic transformation. Using layer-wise activations, we estimate operation-level subspaces via SVD and test their causal relevance through activation steering in four open-weight decoder models. Transformation effects are decodable with $84.8$-$99\%$ accuracy and occupy partially distinct but overlapping subspaces, exceeding random-subspace baselines. Steering experiments show that these directions causally influence predictions, though steerability varies across models; cross-operation steering further reveals structured interference and a dissociation between subspace selectivity and cross-operation independence. These findings indicate that the models encode not only that a hypothesis relates to a premise but also, in part, how it does so, implying that mechanistic analysis and control should operate at the level of semantic operations rather than predicted labels alone.
Code (0)
등록된 구현이 없습니다.
Tasks
Natural Language InferenceSimilar Papers 제목 키워드 기반
Workflow as Knowledge: Semantic Persistence for LLM-Mediated Workflows
Large language model (LLM) applications increasingly use explicit workflows for tool use, retrieval, branching, checkpointing, and human approval. Existing workflow systems already address many execution concerns. This p…
The role of System 1 and System 2 semantic memory structure in human and LLM biases
Implicit biases in both humans and large language models (LLMs) pose significant societal risks. Dual process theories propose that biases arise primarily from associative System 1 thinking, while deliberative System 2 t…
Large Language Models for Zero-shot Inference of Causal Structures in Biology
Genes, proteins and other biological entities influence one another via causal molecular networks. Causal relationships in such networks are mediated by complex and diverse mechanisms, through latent variables, and are o…
Articlesscientific discoveryThe Pragmatic Persona: Discovering LLM Persona through Bridging Inference
Large Language Models (LLMs) reveal inherent and distinctive personas through dialogue. However, most existing persona discovery approaches rely on surface-level lexical or stylistic cues, treating dialogue as a flat seq…
Knowledge GraphsBeyond Hooking Onto the World: Referential Profiles and the Numerical Structure of LLM Grounding
This paper revisits the grounding problem for large language models in light of recent vector-grounding accounts. I accept the shift from classical symbol grounding to vector grounding, but argue that the current debate …