Mitigating Copy Bias in In-Context Learning through Neuron Pruning
원문 초록
Large language models (LLMs) have demonstrated impressive few-shot in-context learning (ICL) abilities. Still, we show that they are sometimes prone to a `copying bias', where they copy answers from provided examples instead of learning the underlying patterns. In this work, we propose a novel and simple method to mitigate such copying bias. First, we create a synthetic task and use the Integrated Gradients method to identify neurons that prioritize copying over generalization. We demonstrate that pruning these neurons consistently improves performance across a diverse set of ICL tasks. We also show that our method is applicable across various LLM architectures, including Transformers and State-Space Models, without requiring modifications. In our analysis, we adopt a task-recognition perspective on ICL and examine task vectors (Hendel et al., 2023) induced by the model. We find that pruning enhances the quality of these vectors, suggesting that the pruned neurons previously hindered effective task recognition.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learning
In-Context Learning for Gradient-Free Receiver Adaptation: Principles, Applications, and Theory
Turbo-ICL: In-Context Learning-Based Turbo Equalization
Assay2Mol: large language model-based drug design using BioAssay context
Optimal Sensor Scheduling and Selection for Continuous-Discrete Kalman Filtering with Auxiliary Dynamics
Precision Spatio-Temporal Feature Fusion for Robust Remote Sensing Change Detection
토론
이 논문과 연결된 포럼 글이 아직 없습니다.