7.8k citations · 21.1k across the 144 of their papers we have counts for
24 papers · 1 filter
Torque-Aware Momentum
Pranshu Malviya, Goncalo Mordido, Aristide Baratin +4
Efficiently exploring complex loss landscapes is key to the performance of deep neural networks. While momentum-based optimizers are widely used in state-of-the-art setups, classic…
TRecViT: A Recurrent Video Transformer
Viorica Pătrăucean, Xu Owen He, Joseph Heyward +10
We propose a novel block for \emph{causal} video modelling. It relies on a time-space-channel factorisation with dedicated blocks for each dimension: gated linear recurrent units (…
Non-Stationary Learning of Neural Networks with Automatic Soft Parameter Reset
Alexandre Galashov, Michalis K. Titsias, András György +4
Neural networks are traditionally trained under the assumption that data come from a stationary distribution. However, settings which violate this assumption are becoming more popu…
A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks
Thomas Schmied, Thomas Adler, Vihang Patil +6
In recent years, there has been a trend in the field of Reinforcement Learning (RL) towards large action models trained offline on large-scale datasets via sequence modeling. Exist…
Retrieval-Augmented Decision Transformer: External Memory for In-context RL
Thomas Schmied, Fabian Paischer, Vihang Patil +3
In-context learning (ICL) is the ability of a model to learn a new task by observing a few exemplars in its context. While prevalent in NLP, this capability has recently also been…
Round and Round We Go! What makes Rotary Positional Encodings useful?
Federico Barbero, Alex Vitvitskyi, Christos Perivolaropoulos +2
Positional Encodings (PEs) are a critical component of Transformer-based Large Language Models (LLMs), providing the attention mechanism with important sequence-position informatio…
Softmax is not Enough (for Sharp Size Generalisation)
Petar Veličković, Christos Perivolaropoulos, Federico Barbero +1
A key property of reasoning systems is the ability to make sharp decisions on their input data. For contemporary AI systems, a key carrier of sharp behaviour is the softmax functio…
Investigating Low-Rank Training in Transformer Language Models: Efficiency and Scaling Analysis
Xiuying Wei, Skander Moalla, Razvan Pascanu +1
State-of-the-art LLMs often rely on scale with high computational costs, which has sparked a research agenda to reduce parameter counts and costs without significantly impacting pe…
When can transformers compositionally generalize in-context?
Seijin Kobayashi, Simon Schug, Yassir Akram +5
Many tasks can be composed from a few independent components. This gives rise to a combinatorial explosion of possible tasks, only some of which might be encountered during trainin…
Normalization and effective learning rates in reinforcement learning
Clare Lyle, Zeyu Zheng, Khimya Khetarpal +4
Normalization layers have recently experienced a renaissance in the deep reinforcement learning and continual learning literature, with several works highlighting diverse benefits…
Transformers meet Neural Algorithmic Reasoners
Wilfried Bounsi, Borja Ibarz, Andrew Dudzik +5
Transformers have revolutionized machine learning with their simple yet effective architecture. Pre-training Transformers on massive text datasets from the Internet has led to unma…
State Soup: In-Context Skill Learning, Retrieval and Mixing
Maciej Pióro, Maciej Wołczyk, Razvan Pascanu +2
A new breed of gated-linear recurrent neural networks has reached state-of-the-art performance on a range of sequence modeling problems. Such models naturally handle long sequences…
Building on Efficient Foundations: Effectively Training LLMs with Structured Feedforward Layers
Xiuying Wei, Skander Moalla, Razvan Pascanu +1
State-of-the-art results in large language models (LLMs) often rely on scale, which becomes computationally expensive. This has sparked a research agenda to reduce these models' pa…
Attention as a Hypernetwork
Simon Schug, Seijin Kobayashi, Yassir Akram +2
Transformers can under some circumstances generalize to novel problem instances whose constituent parts might have been encountered during training, but whose compositions have not…
Transformers need glasses! Information over-squashing in language tasks
Federico Barbero, Andrea Banino, Steven Kapturowski +5
We study how information propagates in decoder-only Transformers, which are the architectural backbone of most existing frontier large language models (LLMs). We rely on a theoreti…
Deep Grokking: Would Deep Neural Networks Generalize Better?
Simin Fan, Razvan Pascanu, Martin Jaggi
Recent research on the grokking phenomenon has illuminated the intricacies of neural networks' training dynamics and their generalization behaviors. Grokking refers to a sharp rise…
No Representation, No Trust: Connecting Representation, Collapse, and Trust Issues in PPO
Skander Moalla, Andrea Miele, Daniil Pyatko +2
Reinforcement learning (RL) is inherently rife with non-stationarity since the states and rewards the agent observes during training depend on its changing policy. Therefore, netwo…
RecurrentGemma: Moving Past Transformers for Efficient Open Language Models
Aleksandar Botev, Soham De, Samuel L Smith +59
We introduce RecurrentGemma, a family of open language models which uses Google's novel Griffin architecture. Griffin combines linear recurrences with local attention to achieve ex…
Revisiting Dynamic Evaluation: Online Adaptation for Large Language Models
Amal Rannen-Triki, Jorg Bornschein, Razvan Pascanu +5
We consider the problem of online fine tuning the parameters of a language model at test time, also known as dynamic evaluation. While it is generally known that this approach impr…
Maxwell's Demon at Work: Efficient Pruning by Leveraging Saturation of Neurons
Simon Dufort-Labbé, Pierluca D'Oro, Evgenii Nikishin +3
When training neural networks, dying neurons -- units becoming inactive or saturated -- are traditionally seen as harmful. This paper sheds new light on this phenomenon. By explori…
Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models
Soham De, Samuel L. Smith, Anushan Fernando +14
Recurrent neural networks (RNNs) have fast inference and scale efficiently on long sequences, but they are difficult to train and hard to scale. We propose Hawk, an RNN with gated…
Disentangling the Causes of Plasticity Loss in Neural Networks
Clare Lyle, Zeyu Zheng, Khimya Khetarpal +4
Underpinning the past decades of work on the design, initialization, and optimization of neural networks is a seemingly innocuous assumption: that the network is trained on a \text…
Fine-tuning Reinforcement Learning Models is Secretly a Forgetting Mitigation Problem
Maciej Wołczyk, Bartłomiej Cupiał, Mateusz Ostaszewski +5
Fine-tuning is a widespread technique that allows practitioners to transfer pre-trained capabilities, as recently showcased by the successful applications of foundation models. How…
Improving fine-grained understanding in image-text pre-training
Ioana Bica, Anastasija Ilić, Matthias Bauer +8
We introduce SPARse Fine-grained Contrastive Alignment (SPARC), a simple method for pretraining more fine-grained multimodal representations from image-text pairs. Given that multi…