김지우 박사과정 EMNLP 2026 Findings 국제 학술대회 논문 채택

26 Aug 2026

DIAL 연구실 소속 인공지능학과 김지우(박사과정) 학생, 이종욱(교신저자) 교수가 참여한 논문 "Causally Redundant Reasoning: A Dual-Disruption Test Framework for Measuring Shortcut Tendencies in Chain-of-Thought"이 자연어처리 분야 최우수 국제 학술대회인 The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026 Findings)에 게재 승인되었으며 오는 10월에 발표될 예정입니다.

Abstract

A chain-of-thought (CoT) rationale can be coherent and answer-consistent while remaining causally unused. We call this phenomenon shortcut reasoning: the model appears to reason through CoT, but its answer does not depend on the expressed trace. We formalize shortcut reasoning as causal redundancy and propose a dual-disruption framework for measuring shortcut tendencies in CoT by testing two complementary aspects of causal explanation: necessity and sufficiency. We operationalize them as two behavioral signals: whether the answer changes under CoT disruption, and whether the answer can be recovered from the CoT after masking the original input. We instantiate this framework with CORD (Causal Redundancy of Reasoning via Disruption), an instance-level score that combines both signals. The two axes reduce complementary ambiguities in single-axis tests: answer preservation under CoT disruption can reflect shortcut use or redundant truthful reasoning, while early answer recovery can reflect shortcut use or a compact truthful rationale. Across 6 models and 2 datasets, CORD reveals a training-paradigm effect: R1-Distill models exhibit roughly twice the shortcut tendency of non-distilled models, while the two axes capture related but distinct signals. Together, these results show that shortcut reasoning can remain hidden behind correct answers and single-axis perturbations, motivating joint tests of necessity and sufficiency.