Publications

Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers

Jonas Ngnaw'e

Sabyasachi Sahoo

Yann Batiste Pequignot

Fr'ed'eric Precioso

Christian Gagné

Despite extensive research on adversarial training strategies to improve robustness, the decisions of even the most robust deep learning mod… (see more)els can still be quite sensitive to imperceptible perturbations, creating serious risks when deploying them for high-stakes real-world applications. While detecting such cases may be critical, evaluating a model's vulnerability at a per-instance level using adversarial attacks is computationally too intensive and unsuitable for real-time deployment scenarios. The input space margin is the exact score to detect non-robust samples and is intractable for deep neural networks. This paper introduces the concept of margin consistency -- a property that links the input space margins and the logit margins in robust models -- for efficient detection of vulnerable samples. First, we establish that margin consistency is a necessary and sufficient condition to use a model's logit margin as a score for identifying non-robust samples. Next, through comprehensive empirical analysis of various robustly trained models on CIFAR10 and CIFAR100 datasets, we show that they indicate strong margin consistency with a strong correlation between their input space margins and the logit margins. Then, we show that we can effectively use the logit margin to confidently detect brittle decisions with such models and accurately estimate robust accuracy on an arbitrarily large test set by estimating the input margins only on a small subset. Finally, we address cases where the model is not sufficiently margin-consistent by learning a pseudo-margin from the feature representation. Our findings highlight the potential of leveraging deep representations to efficiently assess adversarial vulnerability in deployment scenarios.

2024-06-26

ArXiv (preprint)

Implicit Diffusion: Efficient Optimization through Stochastic Sampling

Pierre Marion

Anna Korba

Peter Bartlett

Mathieu Blondel

Valentin De Bortoli

Arnaud Doucet

Felipe Llinares-López

Courtney Paquette

Quentin Berthet

2024-06-26

ICML.cc/2024/Workshop/Differentiable_Almost_Everything (published)

Learning to Design Data-structures: A Case Study of Nearest Neighbor Search

Omar Salemohamed

Vatsal Sharan

Shivam Garg

Laurent Charlin

Gregory Valiant

We propose a general framework for automating data-structure design and apply it to the problem of nearest neighbor search. Our model adapts… (see more) to the underlying data distribution and provides fine-grained control over query and space complexity, enabling the discovery of solutions tailored to problem-specific constraints. We are able to reverse-engineer learned algorithms in several settings. In 1D, the model discovers optimal distribution (in)dependent algorithms such as binary search and variants of interpolation search. In higher dimensions, the model learns solutions that resemble K-d trees in some regimes, while in others, have elements of locality-sensitive hashing.

2024-06-26

ICML.cc/2024/Workshop/Differentiable_Almost_Everything (published)

Gintare Karolina Dziugaite

Mixture of Experts in a Mixture of RL settings

Timon Willi

Johan Samir Obando Ceron

Jakob Nicolaus Foerster

Pablo Samuel Castro

Mixtures of Experts (MoEs) have gained prominence in (self-)supervised learning due to their enhanced inference efficiency, adaptability to … (see more)distributed training, and modularity. Previous research has illustrated that MoEs can significantly boost Deep Reinforcement Learning (DRL) performance by expanding the network's parameter count while reducing dormant neurons, thereby enhancing the model's learning capacity and ability to deal with non-stationarity. In this work, we shed more light on MoEs' ability to deal with non-stationarity and investigate MoEs in DRL settings with"amplified"non-stationarity via multi-task training, providing further evidence that MoEs improve learning capacity. In contrast to previous work, our multi-task results allow us to better understand the underlying causes for the beneficial effect of MoE in DRL training, the impact of the various MoE components, and insights into how best to incorporate them in actor-critic-based DRL networks. Finally, we also confirm results from previous work.

2024-06-26

ArXiv (preprint)

Multimodal foundation world models for generalist embodied agents

Pietro Mazzaglia

Tim Verbelen

Bart Dhoedt

Aaron Courville

Sai Rajeswar

2024-06-26

ArXiv (preprint)

BayTTA: Uncertainty-aware medical image classification with optimized test-time augmentation using Bayesian model averaging

Zeinab Sherkatghanad

Moloud Abdar

Mohammadreza Bakhtyari

Vladimir Makarenkov

Test-time augmentation (TTA) is a well-known technique employed during the testing phase of computer vision tasks. It involves aggregating m… (see more)ultiple augmented versions of input data. Combining predictions using a simple average formulation is a common and straightforward approach after performing TTA. This paper introduces a novel framework for optimizing TTA, called BayTTA (Bayesian-based TTA), which is based on Bayesian Model Averaging (BMA). First, we generate a model list associated with different variations of the input data created through TTA. Then, we use BMA to combine model predictions weighted by their respective posterior probabilities. Such an approach allows one to take into account model uncertainty, and thus to enhance the predictive performance of the related machine learning or deep learning model. We evaluate the performance of BayTTA on various public data, including three medical image datasets comprising skin cancer, breast cancer, and chest X-ray images and two well-known gene editing datasets, CRISPOR and GUIDE-seq. Our experimental results indicate that BayTTA can be effectively integrated into state-of-the-art deep learning models used in medical image analysis as well as into some popular pre-trained CNN models such as VGG-16, MobileNetV2, DenseNet201, ResNet152V2, and InceptionRes-NetV2, leading to the enhancement in their accuracy and robustness performance.

2024-06-25

ArXiv (preprint)

On the consistency of hyper-parameter selection in value-based deep reinforcement learning

Johan Samir Obando Ceron

J. G. Ara'ujo

Aaron Courville

Pablo Samuel Castro

Deep reinforcement learning (deep RL) has achieved tremendous success on various domains through a combination of algorithmic design and car… (see more)eful selection of hyper-parameters. Algorithmic improvements are often the result of iterative enhancements built upon prior approaches, while hyper-parameter choices are typically inherited from previous methods or fine-tuned specifically for the proposed technique. Despite their crucial impact on performance, hyper-parameter choices are frequently overshadowed by algorithmic advancements. This paper conducts an extensive empirical study focusing on the reliability of hyper-parameter selection for value-based deep reinforcement learning agents, including the introduction of a new score to quantify the consistency and reliability of various hyper-parameters. Our findings not only help establish which hyper-parameters are most critical to tune, but also help clarify which tunings remain consistent across different training regimes.

2024-06-25

ArXiv (preprint)

Controlling Large Language Model Agents with Entropic Activation Steering

Nathan Rahn

Pierluca D'Oro

Marc Gendron-Bellemare

The generality of pretrained large language models (LLMs) has prompted increasing interest in their use as in-context learning agents. To be… (see more) successful, such agents must form beliefs about how to achieve their goals based on limited interaction with their environment, resulting in uncertainty about the best action to take at each step. In this paper, we study how LLM agents form and act on these beliefs by conducting experiments in controlled sequential decision-making tasks. To begin, we find that LLM agents are overconfident: They draw strong conclusions about what to do based on insufficient evidence, resulting in inadequately explorative behavior. We dig deeper into this phenomenon and show how it emerges from a collapse in the entropy of the action distribution implied by sampling from the LLM. We then demonstrate that existing token-level sampling techniques are by themselves insufficient to make the agent explore more. Motivated by this fact, we introduce Entropic Activation Steering (EAST), an activation steering method for in-context LLM agents. EAST computes a steering vector as an entropy-weighted combination of representations, and uses it to manipulate an LLM agent's uncertainty over actions by intervening on its activations during the forward pass. We show that EAST can reliably increase the entropy in an LLM agent's actions, causing more explorative behavior to emerge. Finally, EAST modifies the subjective uncertainty an LLM agent expresses, paving the way to interpreting and controlling how LLM agents represent uncertainty about their decisions.

2024-06-24

ICML.cc/2024/Workshop/MI (poster)

Interpretability in Action: Exploratory Analysis of VPT, a Minecraft Agent

Karolis Jucys

George Adamopoulos

Mehrab Hamidi

Stephanie Milani

Mohammad Reza Samsami

Artem Zholus

Sonia Joseph

Blake Richards

Irina Rish

Özgür Şimşek

Understanding the mechanisms behind decisions taken by large foundation models in sequential tasks is critical to ensuring that such systems… (see more) operate transparently and safely. However, interpretability methods have not yet been applied extensively to large-scale agents based on reinforcement learning. In this work, we perform exploratory analysis on the Video PreTraining (VPT) Minecraft playing agent, one of the largest open-source vision-based agents. We try to illuminate its reasoning mechanisms by applying various interpretability techniques. First, we analyze the attention mechanism while the agent solves its training task --- crafting a diamond pickaxe. The agent seems to pay attention to the 4 last frames and several key-frames further back. This provides clues as to how it maintains coherence in the task that takes 3-10 minutes, despite the agent's short memory span of only six seconds. Second, we perform various interventions, which help us uncover a worrying case of goal misgeneralization: VPT mistakenly identifies a villager wearing brown clothes as a tree trunk and punches it to death, when positioned stationary under green tree leaves. We demonstrate similar misbehavior in a related agent (STEVE-1), which motivates the use of VPT as a model organism for large-scale vision-based agent interpretability.

2024-06-24

ICML.cc/2024/Workshop/MI (poster)

Gintare Karolina Dziugaite

Robust Unlearning via Mechanistic Localizations

Phillip Huang Guo

Aaquib Syed

Abhay Sheshadri

Aidan Ewart

Methods for machine unlearning in large language models seek to remove undesirable knowledge or capabilities without compromising general la… (see more)nguage modeling performance. This work investigates the use of mechanistic interpretability to improve the precision and effectiveness of unlearning. We demonstrate that localizing unlearning to components with particular mechanisms in factual recall leads to more robust unlearning across different input/output formats, relearning, and latent knowledge, and reduces unintended side effects compared to nonlocalized unlearning. Additionally, we analyze the strengths and weaknesses of different automated (rather than manual) interpretability methods for guiding unlearning, finding that their corresponding unlearned models require smaller edit sizes to achieve unlearning but are much less robust.

2024-06-24

ICML.cc/2024/Workshop/MI (spotlight)

The Responsible Foundation Model Development Cheatsheet: A Review of Tools&Resources

Shayne Longpre

Stella Biderman

Alon Albalak

Hailey Schoelkopf

Daniel McDuff

Sayash Kapoor

Kevin Klyman

Kyle Lo

Gabriel Ilharco

Nay San

Maribeth Rauh

Aviya Skowron

Bertie Vidgen

Laura Weidinger

Arvind Narayanan

Victor Sanh

David Ifeoluwa Adelani

Percy Liang

Rishi Bommasani

Peter Henderson 0002 … (see 3 more)

Sasha Luccioni

Yacine Jernite

Luca Soldaini

2024-06-24

ArXiv (preprint)