1.1k citations · 1.8k across the 50 of their papers we have counts for
22 papers · 1 filter
Conformalized Large Language Models under Configuration Shift
Yuqicheng Zhu, Jialin Yu, Lin Li +7
Conformal prediction (CP) is a distribution-free framework for uncertainty quantification that has recently been adapted to large language models (LLMs), providing prediction sets…
Mixture of Experts Made Intrinsically Interpretable
Xingyi Yang, Constantin Venhoff, Ashkan Khakzar +4
Neurons in large language models often exhibit \emph{polysemanticity}, simultaneously encoding multiple unrelated concepts and obscuring interpretability. Instead of relying on pos…
What Makes and Breaks Safety Fine-tuning? A Mechanistic Study
Samyak Jain, Ekdeep Singh Lubana, Kemal Oksuz +4
Safety fine-tuning helps align Large Language Models (LLMs) with human preferences for their safe deployment. To better understand the underlying factors that make models safe via…
Fine-tuning can cripple your foundation model; preserving features may be the solution
Jishnu Mukhoti, Yarin Gal, Philip H. S. Torr +1
Pre-trained foundation models, due to their enormous capacity and exposure to vast amounts of data during pre-training, are known to have learned plenty of real-world concepts. An…
Graph Inductive Biases in Transformers without Message Passing
Liheng Ma, Chen Lin, Derek Lim +5
Transformers for graph data are increasingly widely studied and successful in numerous learning tasks. Graph inductive biases are crucial for Graph Transformers, and previous works…
Computationally Budgeted Continual Learning: What Does Matter?
Ameya Prabhu, Hasan Abed Al Kader Hammoud, Puneet Dokania +4
Continual Learning (CL) aims to sequentially train models on streams of incoming data that vary in distribution by preserving previous knowledge while adapting to new data. Current…
RegMixup: Mixup as a Regularizer Can Surprisingly Improve Accuracy and Out Distribution Robustness
Francesco Pinto, Harry Yang, Ser-Nam Lim +2
We show that the effectiveness of the well celebrated Mixup [Zhang et al., 2018] can be further improved if instead of using it as the sole learning objective, it is utilized as an…
Catastrophic overfitting can be induced with discriminative non-robust features
Guillermo Ortiz-Jiménez, Pau de Jorge, Amartya Sanyal +5
Adversarial training (AT) is the de facto method for building robust neural networks, but it can be computationally expensive. To mitigate this, fast single-step attacks can be use…
Make Some Noise: Reliable and Efficient Single-Step Adversarial Training
Pau de Jorge, Adel Bibi, Riccardo Volpi +4
Recently, Wong et al. showed that adversarial training with single-step FGSM leads to a characteristic failure mode named Catastrophic Overfitting (CO), in which a model becomes su…
A Step Towards Efficient Evaluation of Complex Perception Tasks in Simulation
Jonathan Sadeghi, Blaine Rogers, James Gunn +4
There has been increasing interest in characterising the error behaviour of systems which contain deep learning models before deploying them into any safety-critical scenario. Howe…
ANCER: Anisotropic Certification via Sample-wise Volume Maximization
Francisco Eiras, Motasem Alfarra, M. Pawan Kumar +4
Randomized smoothing has recently emerged as an effective tool that enables certification of deep neural network classifiers at scale. All prior art on randomized smoothing has foc…
No Cost Likelihood Manipulation at Test Time for Making Better Mistakes in Deep Networks
Shyamgopal Karthik, Ameya Prabhu, Puneet K. Dokania +1
There has been increasing interest in building deep hierarchy-aware classifiers that aim to quantify and reduce the severity of mistakes, and not just reduce the number of errors.…
On Batch Normalisation for Approximate Bayesian Inference
Jishnu Mukhoti, Puneet K. Dokania, Philip H. S. Torr +1
We study batch normalisation in the context of variational inference methods in Bayesian neural networks, such as mean-field or MC Dropout. We show that batch-normalisation does no…
Continual Learning in Low-rank Orthogonal Subspaces
Arslan Chaudhry, Naeemullah Khan, Puneet K. Dokania +1
In continual learning (CL), a learner is faced with a sequence of tasks, arriving one after the other, and the goal is to remember all the tasks once the continual learning experie…
How benign is benign overfitting?
Amartya Sanyal, Puneet K Dokania, Varun Kanade +1
We investigate two causes for adversarial vulnerability in deep neural networks: bad data and (poorly) trained models. When trained with SGD, deep neural networks essentially achie…
Simulation-Based Inference for Global Health Decisions
Christian Schroeder de Witt, Bradley Gram-Hansen, Nantas Nardelli +8
The COVID-19 pandemic has highlighted the importance of in-silico epidemiological modelling in predicting the dynamics of infectious diseases to inform health policy and decision m…
Calibrating Deep Neural Networks using Focal Loss
Jishnu Mukhoti, Viveka Kulharia, Amartya Sanyal +3
Miscalibration - a mismatch between a model's confidence and its correctness - of Deep Neural Networks (DNNs) makes their predictions hard to rely on. Ideally, we want networks to…
Using Hindsight to Anchor Past Knowledge in Continual Learning
Arslan Chaudhry, Albert Gordo, Puneet K. Dokania +2
In continual learning, the learner faces a stream of data whose distribution changes over time. Modern neural networks are known to suffer under this setting, as they quickly forge…
Mirror Descent View for Neural Network Quantization
Thalaiyasingam Ajanthan, Kartik Gupta, Philip H. S. Torr +2
Quantizing large Neural Networks (NN) while maintaining the performance is highly desirable for resource-limited devices due to reduced memory and time complexity. It is usually fo…
On Tiny Episodic Memories in Continual Learning
Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny +4
In continual learning (CL), an agent learns from a stream of tasks leveraging prior experience to transfer knowledge to future tasks. It is an ideal framework to decrease the amoun…
Robustness via Deep Low-Rank Representations
Amartya Sanyal, Varun Kanade, Philip H. S. Torr +1
We investigate the effect of the dimensionality of the representations learned in Deep Neural Networks (DNNs) on their robustness to input perturbations, both adversarial and rando…
Minding the Gaps for Block Frank-Wolfe Optimization of Structured SVMs
Anton Osokin, Jean-Baptiste Alayrac, Isabella Lukasewitz +2
In this paper, we propose several improvements on the block-coordinate Frank-Wolfe (BCFW) algorithm from Lacoste-Julien et al. (2013) recently used to optimize the structured suppo…