Sense Representations Are Inducible Interfaces

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cruz, Jan Christian Blaise, Aji, Alham Fikri
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910266998915072
author Cruz, Jan Christian Blaise
Aji, Alham Fikri
author_facet Cruz, Jan Christian Blaise
Aji, Alham Fikri
contents Sense representations (explicit, per-token meaning decompositions) are useful for disambiguation, steering, and cross-lingual alignment, but existing approaches require models to be pretrained with sense structure baked in. We introduce ACROS, which induces an explicit sense pathway into a frozen pretrained decoder LM through a gated residual addition. On SmolLM2-360M, ACROS preserves base LM quality while supporting three uses of the same induced variables: zero-shot word-sense disambiguation (64.95 F1 on Raganato ALL, competitive with the WordNet first-sense heuristic), low-KL lexical steering across 5,161 CoInCo cases where a simple non-oracle proxy recovers about 90% of positive shifts, and SENSIA cross-lingual adaptation to four languages (mean R@1 0.988, target FLORES PPL 7.94). ACROS makes sense representations an inducible interface for ordinary pretrained LMs.
format Preprint
id arxiv_https___arxiv_org_abs_2605_28669
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Sense Representations Are Inducible Interfaces
Cruz, Jan Christian Blaise
Aji, Alham Fikri
Computation and Language
Artificial Intelligence
Sense representations (explicit, per-token meaning decompositions) are useful for disambiguation, steering, and cross-lingual alignment, but existing approaches require models to be pretrained with sense structure baked in. We introduce ACROS, which induces an explicit sense pathway into a frozen pretrained decoder LM through a gated residual addition. On SmolLM2-360M, ACROS preserves base LM quality while supporting three uses of the same induced variables: zero-shot word-sense disambiguation (64.95 F1 on Raganato ALL, competitive with the WordNet first-sense heuristic), low-KL lexical steering across 5,161 CoInCo cases where a simple non-oracle proxy recovers about 90% of positive shifts, and SENSIA cross-lingual adaptation to four languages (mean R@1 0.988, target FLORES PPL 7.94). ACROS makes sense representations an inducible interface for ordinary pretrained LMs.
title Sense Representations Are Inducible Interfaces
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2605.28669