Razvan Pascanu

Murray P. Shanahan

James L McClelland

2025-05-01

arXiv (publié)

On the generalization of language models from in-context learning and finetuning: a controlled study

Andrew Lampinen

Arslan Chaudhry

Stephanie C.Y. Chan

Cody Wild

Diane Wan

Alexander Y. Ku

Jorg Bornschein

Murray P. Shanahan

James L McClelland

Large language models exhibit exciting capabilities, yet can show surprisingly narrow generalization from finetuning. E.g. they can fail to … (voir plus)generalize to simple reversals of relations they are trained on, or fail to make simple logical deductions based on trained information. These failures to generalize from fine-tuning can hinder practical application of these models. On the other hand, language models' in-context learning shows different inductive biases, and can generalize better in some cases. Here, we explore these differences in generalization between in-context- and fine-tuning-based learning. To do so, we constructed several novel datasets to evaluate and improve models' abilities to generalize from finetuning data. The datasets are designed to create clean tests of generalization, by isolating the knowledge in the dataset from that in pretraining. We expose pretrained large models to controlled subsets of the information in these datasets -- either in context, or through fine-tuning -- and evaluate their performance on test sets that require various types of generalization. We find overall that in data-matched settings, in-context learning can generalize more flexibly than fine-tuning (though we also find some qualifications of prior findings, such as cases when fine-tuning can generalize to reversals embedded in a larger structure of knowledge). We build on these findings to propose a method to enable improved generalization from fine-tuning: adding in-context inferences to finetuning data. We show that this method improves generalization across various splits of our datasets and other benchmarks. Our results have implications for understanding the inductive biases of different modes of learning in language models, and practically improving their performance.

2025-05-01

ArXiv (prépublication)

LLMs are Greedy Agents: Effects of RL Fine-tuning on Decision-Making Abilities

Thomas Schmied

Jorg Bornschein

Jordi Grau-Moya

Markus Wulfmeier

2025-04-22

ArXiv (prépublication)

Why do LLMs attend to the first token?

Federico Barbero

'Alvaro Arroyo

Xiangming Gu

Christos Perivolaropoulos

Michael M. Bronstein

Petar Velivckovi 'c

2025-04-03

ArXiv (prépublication)

LLMs are Greedy Agents: Effects of RL Fine-tuning on Decision-Making Abilities

Thomas Schmied

Jorg Bornschein

Jordi Grau-Moya

Markus Wulfmeier

2025-04-01

arXiv (publié)

Why do LLMs attend to the first token?

Federico Barbero

'Alvaro Arroyo

Xiangming Gu

Christos Perivolaropoulos

Michael M. Bronstein

Petar Veličković

2025-04-01

arXiv (publié)

NoProp: Training Neural Networks without Back-propagation or Forward-propagation

Qinyu Li

Yee Whye Teh

2025-03-31

ArXiv (prépublication)

NoProp: Training Neural Networks without Back-propagation or Forward-propagation

Qinyu Li

Yee Whye Teh

2025-03-31

ArXiv (prépublication)

NoProp: Training Neural Networks without Full Back-propagation or Full Forward-propagation

Qinyu Li

Yee Whye Teh

The canonical deep learning approach for learning requires computing a gradient term at each block by back-propagating the error signal from… (voir plus) the output towards each learnable parameter. Given the stacked structure of neural networks, where each block builds on the representation of the block below, this approach leads to hierarchical representations. More abstract features live on the top blocks of the model, while features on lower blocks are expected to be less abstract. In contrast to this, we introduce a new learning method named NoProp, which does not rely on either forward or backwards propagation across the entire network. Instead, NoProp takes inspiration from diffusion and flow matching methods, where each block independently learns to denoise a noisy target using only local targets and back-propagation within the block. We believe this work takes a first step towards introducing a new family of learning methods that does not learn hierarchical representations -- at least not in the usual sense. NoProp needs to fix the representation at each block beforehand to a noised version of the target, learning a local denoising process that can then be exploited at inference. We demonstrate the effectiveness of our method on MNIST, CIFAR-10, and CIFAR-100 image classification benchmarks. Our results show that NoProp is a viable learning algorithm, is easy to use and computationally efficient. By departing from the traditional learning paradigm which requires back-propagating a global error signal, NoProp alters how credit assignment is done within the network, enabling more efficient distributed learning as well as potentially impacting other characteristics of the learning process.

2025-03-31

ArXiv (prépublication)

How do language models learn facts? Dynamics, curricula and hallucinations

Nicolas Zucchet

Jorg Bornschein

Stephanie Chan

Andrew Lampinen

Soham De

2025-03-27

ArXiv (prépublication)

How do language models learn facts? Dynamics, curricula and hallucinations

Nicolas Zucchet

Jorg Bornschein

Stephanie Chan

Andrew Lampinen

Soham De

2025-03-27

ArXiv (prépublication)

From Markov to Laplace: How Mamba In-Context Learns Markov Chains

Marco Bondaschi

Nived Rajaraman

Xiuying Wei

Kannan Ramchandran

Caglar Gulcehre

Michael C. Gastpar

Ashok Vardhan Makkuva

2025-02-14

ArXiv (prépublication)