Mohammad Pezeshki

Sarath Chandar

Pascal Vincent

2025-07-07

colmweb.org/COLM/2025/Conference (accepté)

Compositional Risk Minimization

Charles Arnal

Kartik Ahuja

2025-05-01

ICML.cc/2025/Conference (poster)

Improving the Scaling Laws of Synthetic Data with Deliberate Practice

Elvis Dohmatob

Pietro Astolfi

Melissa Hall

Jakob Verbeek

Michal Drozdzal

2025-05-01

ICML.cc/2025/Conference (présentation orale)

Steering Large Language Model Activations in Sparse Spaces

Reza Bayat

Ali Rahimi-Kalahroudi

Sarath Chandar

Pascal Vincent

A key challenge in AI alignment is guiding large language models (LLMs) to follow desired behaviors at test time. Activation steering, which… (voir plus) modifies internal model activations during inference, offers a potential solution. However, prior work in dense activation spaces struggles with superposition, wherein multiple features become entangled, limiting interpretability and precise control. In contrast, sparse representations provide an untapped opportunity for more interpretable behavior modulation. In this work, we introduce sparse activation steering (SAS), a method that leverages sparse autoencoders (SAEs) to steer LLM behavior in sparse spaces. By isolating behavior-specific features through a contrastive prompt-pairing approach, we define a set of features that can selectively reinforce or suppress behaviors. Experiments on Gemma 2 LLMs show that SAS vectors enable nuanced behavioral modulation and finer-grained control. Furthermore, scaling SAEs improves monosemanticity of SAS vectors, suggesting more reliable and interpretable interventions.

2025-02-28

ArXiv (prépublication)

arxiv.org

Steering Large Language Model Activations in Sparse Spaces

Reza Bayat

Ali Rahimi-Kalahroudi

Sarath Chandar

Pascal Vincent

A key challenge in AI alignment is guiding large language models (LLMs) to follow desired behaviors at test time. Activation steering, which… (voir plus) modifies internal model activations during inference, offers a potential solution. However, prior work in dense activation spaces struggles with superposition, wherein multiple features become entangled, limiting interpretability and precise control. In contrast, sparse representations provide an untapped opportunity for more interpretable behavior modulation. In this work, we introduce sparse activation steering (SAS), a method that leverages sparse autoencoders (SAEs) to steer LLM behavior in sparse spaces. By isolating behavior-specific features through a contrastive prompt-pairing approach, we define a set of features that can selectively reinforce or suppress behaviors. Experiments on Gemma 2 LLMs show that SAS vectors enable nuanced behavioral modulation and finer-grained control. Furthermore, scaling SAEs improves monosemanticity of SAS vectors, suggesting more reliable and interpretable interventions.

2025-02-28

ArXiv (prépublication)

arxiv.org

Improving the Scaling Laws of Synthetic Data with Deliberate Practice

Elvis Dohmatob

Pietro Astolfi

Melissa Hall

Jakob Verbeek

Michal Drozdzal

Inspired by the principle of deliberate practice in human learning, we propose Deliberate Practice for Synthetic Data Generation (DP), a nov… (voir plus)el framework that improves sample efficiency through dynamic synthetic data generation. Prior work has shown that scaling synthetic data is inherently challenging, as naively adding new data leads to diminishing returns. To address this, pruning has been identified as a key mechanism for improving scaling, enabling models to focus on the most informative synthetic samples. Rather than generating a large dataset and pruning it afterward, DP efficiently approximates the direct generation of informative samples. We theoretically show how training on challenging, informative examples improves scaling laws and empirically validate that DP achieves better scaling performance with significantly fewer training samples and iterations. On ImageNet-100, DP generates 3.4x fewer samples and requires six times fewer iterations, while on ImageNet-1k, it generates 8x fewer samples with a 30 percent reduction in iterations, all while achieving superior performance compared to prior work.

2025-02-21

ArXiv (prépublication)

arxiv.org

The Pitfalls of Memorization: When Memorization Hurts Generalization

David Lopez-Paz

Neural networks often learn simple explanations that fit the majority of the data while memorizing exceptions that deviate from these explan… (voir plus)ations.This behavior leads to poor generalization when the learned explanations rely on spurious correlations. In this work, we formalize the interplay between memorization and generalization, showing that spurious correlations would particularly lead to poor generalization when are combined with memorization. Memorization can reduce training loss to zero, leaving no incentive to learn robust, generalizable patterns. To address this, we propose memorization-aware training (MAT), which uses held-out predictions as a signal of memorization to shift a model's logits. MAT encourages learning robust patterns invariant across distributions, improving generalization under distribution shifts.

2025-01-22

ICLR.cc/2025/Conference (poster)

Deliberate Practice with Synthetic Data

Pietro Astolfi

Melissa Hall

Jakob Verbeek

Michal Drozdzal

2024-10-10

NeurIPS.cc/2024/Workshop/AFM (poster)

The Pitfalls of Memorization: When Memorization Hinders Generalization

David Lopez-Paz

Neural networks often learn simple explanations that fit the majority of the data while memorizing exceptions that deviate from these explan… (voir plus)ations. This leads to poor generalization when the learned explanations are spurious. In this work, we formalize

2024-10-10

NeurIPS.cc/2024/Workshop/SciForDL (poster)

The Pitfalls of Memorization: When Memorization Hinders Generalization

David Lopez-Paz

Neural networks often learn simple explanations that fit the majority of the data while memorizing exceptions that deviate from these explan… (voir plus)ations. This leads to poor generalization when the learned explanations are spurious. In this work, we formalize

2024-10-10

NeurIPS.cc/2024/Workshop/SciForDL (poster)

Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok

Tikeng Notsawo Pascal Junior

Pascal Notsawo

2024-03-04

ICLR.cc/2024/Workshop/ME-FoMo (poster)

Feedback-guided Data Synthesis for Imbalanced Classification

Michal Drozdzal

Current status quo in machine learning is to use static datasets of real images for training, which often come from long-tailed distribution… (voir plus)s. With the recent advances in generative models, researchers have started augmenting these static datasets with synthetic data, reporting moderate performance improvements on classification tasks. We hypothesize that these performance gains are limited by the lack of feedback from the classifier to the generative model, which would promote the usefulness of the generated samples to improve the classifier's performance. In this work, we introduce a framework for augmenting static datasets with useful synthetic samples, which leverages one-shot feedback from the classifier to drive the sampling of the generative model. In order for the framework to be effective, we find that the samples must be close to the support of the real data of the task at hand, and be sufficiently diverse. We validate three feedback criteria on a long-tailed dataset (ImageNet-LT) as well as a group-imbalanced dataset (NICO++). On ImageNet-LT, we achieve state-of-the-art results, with over 4 percent improvement on underrepresented classes while being twice efficient in terms of the number of generated synthetic samples. NICO++ also enjoys marked boosts of over 5 percent in worst group accuracy. With these results, our framework paves the path towards effectively leveraging state-of-the-art text-to-image models as data sources that can be queried to improve downstream applications.

2023-10-30

NeurIPS.cc/2023/Workshop/SyntheticData4ML (poster)