In-context learning enables large language models to perform novel tasks through few-shot demonstrations. However, demonstrations per se can naturally contain noise and conflicting examples, making this capability vulnerable. To understand how models process such conflicts, we study demonstration-dependent tasks requiring models to infer underlying patterns, a process we characterize as rule inference. We find that models suffer substantial performance degradation from a single demonstration with corrupted rule. This systematic misleading behavior motivates our investigation of how models process conflicting evidence internally. Using linear probes and logit lens analysis, we discover that under corruption models encode both correct and incorrect rules in intermediate layers but develop prediction confidence only in late layers, revealing a two-phase computational structure. We then identify attention heads for each phase underlying the reasoning failures: Vulnerability Heads in early-to-middle layers exhibit positional attention bias with high sensitivity to corruption, while Susceptible Heads in late layers significantly reduce support for correct predictions when exposed to the corrupted evidence. Targeted ablation validates our findings, with masking a small number of identified heads improving performance by over 10%.
Is Cognition Consistent with Perception? Assessing and Mitigating Multimodal Knowledge Conflicts in Document Understanding
Zhejiang Key Laboratory of Accessible Perception and Intelligent Systems, Zhejiang University（浙江可感知智能系统重点实验室，浙江大学）；Alibaba Group（阿里巴巴集团）；Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and DataSecurity（杭州高新技术区（滨江）区块链与数据安全研究院）
Getting Sick After Seeing a Doctor? Diagnosing and Mitigating Knowledge Conflicts in Event Temporal Reasoning
Hong Kong University of Science and Technology（香港科技大学）；University of Southern California（南加州大学）；Tencent AI Lab（腾讯AI实验室）；University of California, Davis（加州大学戴维斯分校）
Mixture of Demonstrations for Textual Graph Understanding and Question Answering
Independent Researcher, Wayne State University（韦恩州立大学独立研究者）；Institute for Clarity in Documentation（文档清晰研究所）；Inria Paris-Rocquencourt（巴黎-罗克琴court研究所）；Rajiv Gandhi University（拉吉夫·甘地大学）；Tsinghua University（清华大学）；Palmer Research Laboratories（帕勒尔研究实验室）
Meta AI；Allen Institute for AI（艾伦人工智能研究所）；University of Washington（华盛顿大学）
Large language models (LMs) are able to in-context learn -- perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required -- randomly replacing labels in the demonstrations barely hurts performance on a range of classification and multi-choce tasks, consistently over 12 different models including …
The Impact of Demonstrations on Multilingual In-Context Learning: A Multidimensional Analysis
Saarland University（萨尔大学）；Bosch Center for AI（博世人工智能中心）；LMU Munich（慕尼黑大学）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；Eastern Institute of Technology（东方理工学院）；Mila（米拉研究所）；McGill University（麦吉尔大学）
Beihang University（北京航空航天大学）；University of Ottawa（渥太华大学）
In this paper, we propose a test, called Flagged-1-Bit (F1B) test, to study the intrinsic capability of recurrent neural networks in sequence learning. Four different recurrent network models are studied both analytically and experimentally using this test. Our results suggest that in general there exists a conflict between feature selection and feature memorization in sequence learning. Such a conflict can be resolved either using a gating mechanism as in LSTM, or by increasing the state dimension as in Vanilla RNN. Gated models resolve this conflict by adaptively adjusting their state-update…
University of California, Riverside（加州大学河滨分校）；Yale University（耶鲁大学）；Georgia Institute of Technology（佐治亚理工学院）；University of Michigan, Ann Arbor（密歇根大学安娜堡分校）；University of Toronto（多伦多大学）
ETH Zürich（苏黎世联邦理工学院）；MIT（麻省理工学院）；EPFL（洛桑联邦理工学院）；University of Cambridge（剑桥大学）
Understanding the origins of militarized conflict is a complex, yet important undertaking. Existing research seeks to build this understanding by considering bi-lateral relationships between entity pairs (dyadic causes) and multi-lateral relationships among multiple entities (systemic causes). The aim of this work is to compare these two causes in terms of how they correlate with conflict between two entities. We do this by devising a set of textual and graph-based features which represent each of the causes. The features are extracted from Wikipedia and modeled as a large graph. Nodes in this…
