Publications (26)
Neural Networks and the Chomsky Hierarchy
Grégoire Delétang, Anian Ruoss, Jordi Grau-Moya +8
Reliable generalization lies at the heart of safe ML and AI. However, understanding when and how neural networks generalize remains one of the most important unsolved problems in t…
Why is prompting hard? Understanding prompts on binary sequence predictors
Li Kevin Wenliang, Anian Ruoss, Jordi Grau-Moya +2
Frontier models can be prompted or conditioned to do many tasks, but finding good prompts is not always easy, nor is understanding some performant prompts. We view prompting as fin…
Causal Analysis of Agent Behavior for AI Safety
Grégoire Déletang, Jordi Grau-Moya, Miljan Martic +6
As machine learning systems become more powerful they also become increasingly unpredictable and opaque. Yet, finding human-understandable explanations of how they work is essentia…
Randomized Positional Encodings Boost Length Generalization of Transformers
Anian Ruoss, Grégoire Delétang, Tim Genewein +5
Transformers have impressive generalization capabilities on tasks with a fixed context length. However, they fail to generalize to sequences of arbitrary length, even for seemingly…
Understanding Prompt Tuning and In-Context Learning via Meta-Learning
Tim Genewein, Li Kevin Wenliang, Jordi Grau-Moya +3
Prompting is one of the main ways to adapt a pretrained model to target tasks. Besides manually constructing prompts, many prompt optimization methods have been proposed in the lit…
A Nonparametric Conjugate Prior Distribution for the Maximizing Argument of a Noisy Function
Pedro A. Ortega, Jordi Grau-Moya, Tim Genewein +2
We propose a novel Bayesian approach to solve stochastic optimization problems that involve finding extrema of noisy, nonlinear functions. Previous work has focused on representing…
LMAct: A Benchmark for In-Context Imitation Learning with Long Multimodal Demonstrations
Anian Ruoss, Fabio Pardo, Harris Chan +3
In this paper, we present a benchmark to pressure-test today's frontier models' multimodal decision-making capabilities in the very long-context regime (up to one million tokens) a…
Planning with Information-Processing Constraints and Model Uncertainty in Markov Decision Processes
Jordi Grau-Moya, Felix Leibfried, Tim Genewein +1
Information-theoretic principles for learning and acting have been proposed to solve particular classes of Markov Decision Problems. Mathematically, such approaches are governed by…
Compression via Pre-trained Transformers: A Study on Byte-Level Multimodal Data
David Heurtel-Depeiges, Anian Ruoss, Joel Veness +1
Foundation models are strong data compressors, but when accounting for their parameter size, their compression ratios are inferior to standard compression algorithms. Naively reduc…
Algorithms for Causal Reasoning in Probability Trees
Tim Genewein, Tom McGrath, Grégoire Déletang +4
Probability trees are one of the simplest models of causal generative processes. They possess clean semantics and -- unlike causal Bayesian networks -- they can represent context-s…
Learning Universal Predictors
Jordi Grau-Moya, Tim Genewein, Marcus Hutter +8
Meta-learning has emerged as a powerful approach to train neural networks to learn new tasks quickly from limited data. Broad exposure to different tasks leads to versatile represe…
Shaking the foundations: delusions in sequence models for interaction and control
Pedro A. Ortega, Markus Kunesch, Grégoire Delétang +16
The recent phenomenal success of language models has reinvigorated machine learning research, and large sequence models such as transformers are being applied to a variety of domai…
Language Modeling Is Compression
Grégoire Delétang, Anian Ruoss, Paul-Ambroise Duquenne +9
It has long been established that predictive models can be transformed into lossless compressors and vice versa. Incidentally, in recent years, the machine learning community has f…
Abstraction in decision-makers with limited information processing capabilities
Tim Genewein, Daniel A. Braun
A distinctive property of human and animal intelligence is the ability to form abstractions by neglecting irrelevant information which allows to separate structure from noise. From…
Your Policy Regularizer is Secretly an Adversary
Rob Brekelmans, Tim Genewein, Jordi Grau-Moya +4
Policy regularization methods such as maximum entropy regularization are widely used in reinforcement learning to improve the robustness of a learned policy. In this paper, we show…
Sinkhorn AutoEncoders
Giorgio Patrini, Rianne van den Berg, Patrick Forré +5
Optimal transport offers an alternative to maximum likelihood for learning generative autoencoding models. We show that minimizing the p-Wasserstein distance between the generator…
Memory-Based Meta-Learning on Non-Stationary Distributions
Tim Genewein, Grégoire Delétang, Anian Ruoss +7
Memory-based meta-learning is a technique for approximating Bayes-optimal predictors. Under fairly general conditions, minimizing sequential prediction error, measured by the log l…
Amortized Planning with Large-Scale Transformers: A Case Study on Chess
Anian Ruoss, Grégoire Delétang, Sourabh Medapati +7
This paper uses chess, a landmark planning problem in AI, to assess transformers' performance on a planning task where memorization is futile $\unicode{x2013}$ even at a large scal…
From AGI to ASI
Tim Genewein, Matija Franklin, Alexander Lerchner +11
Over the last decade, building human-level artificial general intelligence has moved from far-fetched speculation to being a concrete next-decade target for many of the largest AI…
Meta-trained agents implement Bayes-optimal agents
Vladimir Mikulik, Grégoire Delétang, Tom McGrath +4
Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task within a target distribution. A previous theoretical study has argued that this remar…
Group Pruning using a Bounded-Lp norm for Group Gating and Regularization
Chaithanya Kumar Mummadi, Tim Genewein, Dan Zhang +2
Deep neural networks achieve state-of-the-art results on several tasks while increasing in complexity. It has been shown that neural networks can be pruned during training by impos…
An information-theoretic on-line update principle for perception-action coupling
Zhen Peng, Tim Genewein, Felix Leibfried +1
Inspired by findings of sensorimotor coupling in humans and animals, there has recently been a growing interest in the interaction between action and perception in robotic systems…
On Detecting Adversarial Perturbations
Jan Hendrik Metzen, Tim Genewein, Volker Fischer +1
Machine learning and deep learning in particular has advanced tremendously on perceptual tasks in recent years. However, it remains vulnerable against adversarial perturbations of…
Beyond Bayes-optimality: meta-learning what you know you don't know
Jordi Grau-Moya, Grégoire Delétang, Markus Kunesch +11
Meta-training agents with memory has been shown to culminate in Bayes-optimal agents, which casts Bayes-optimality as the implicit solution to a numerical optimization problem rath…
Model-Free Risk-Sensitive Reinforcement Learning
Grégoire Delétang, Jordi Grau-Moya, Markus Kunesch +4
We extend temporal-difference (TD) learning in order to obtain risk-sensitive, model-free reinforcement learning algorithms. This extension can be regarded as modification of the R…
Meta-learning of Sequential Strategies
Pedro A. Ortega, Jane X. Wang, Mark Rowland +21
In this report we review memory-based meta-learning as a tool for building sample-efficient strategies that learn from past experience to adapt to any task within a target class. O…