paper-with-me

Papers

Why Larger Language Models Do In-context Learning Differently?

2024-05-30 · Zhenmei Shi, Junyi Wei, Zhuoyan Xu, YIngyu Liang

Large language models (LLM) have emerged as a powerful tool for AI, with the key ability of in-context learning (ICL), where they can perform well on unseen tasks based on a brief series of task examples without necessitating any adjustments to the model parameters. One recent interesting mysterious observation is that models of different scales may have different ICL behaviors: larger models tend to be more sensitive to noise in the test context. This work studies this observation theoretically aiming to improve the understanding of LLM and ICL. We analyze two stylized settings: (1) linear regression with one-layer single-head linear transformers and (2) parity classification with two-layer multiple attention heads transformers (non-linear data and non-linear model). In both settings, we give closed-form optimal solutions and find that smaller models emphasize important hidden features while larger ones cover more hidden features; thus, smaller models are more robust to noise while larger ones are more easily distracted, leading to different ICL behaviors. This sheds light on where transformers pay attention to and how that affects ICL. Preliminary experimental results on large base and chat models provide positive support for our analysis.

📄 PDF Abstract BibTeX arXiv:2405.19592

Code (0)

등록된 구현이 없습니다.

Tasks

In-Context Learning

Methods 이 논문이 사용한 방법론

BASE 설명 없음
Linear Regression Linear Regression is a method for modelling a relationship between a dependent variable and independent variables. These models can be fit with numerous approaches. The most…

Similar Papers 제목 키워드 기반

Larger language models do in-context learning differently

2023-03-07 · Jerry Wei, Jason Wei, Yi Tay, Dustin Tran 외

We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across va…

In-Context Learning

Vectors from Larger Language Models Predict Human Reading Time and fMRI Data More Poorly when Dimensionality Expansion is Controlled

2025-05-18 · Yi-Chien Lin, Hongao Zhu, William Schuler

The impressive linguistic abilities of large language models (LLMs) have recommended them as models of human sentence processing, with some conjecturing a positive 'quality-power' relationship (Wilcox et al., 2023), in w…

Sentence

Training Trajectories of Language Models Across Scales

2022-12-19 · Mengzhou Xia, Mikel Artetxe, Chunting Zhou, Xi Victoria Lin 외

Scaling up language models has led to unprecedented performance gains, but little is understood about how the training dynamics change as models get larger. How do language models of different sizes learn during pre-trai…

In-Context LearningMultiple-choice

Transformers generalize differently from information stored in context vs in weights

2022-10-11 · Stephanie C. Y. Chan, Ishita Dasgupta, Junkyung Kim, Dharshan Kumaran 외

Transformer models can use two fundamentally different kinds of information: information stored in weights during training, and information provided ``in-context'' at inference time. In this work, we show that transforme…

In-Context Learning

Solving Cosine Similarity Underestimation between High Frequency Words by L2 Norm Discounting

2023-05-17 · Saeth Wannasuphoprasit, Yi Zhou, Danushka Bollegala

Cosine similarity between two words, computed using their contextualised token embeddings obtained from masked language models (MLMs) such as BERT has shown to underestimate the actual similarity between those words (Zho…

Word Similarity