Nicolas Flammarion

papers

2,453

total citations

papers (21)

Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity

NEURIPS 2021arXiv

118

citations

Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning

ICML 2024arXiv

citations

Gradient flow dynamics of shallow ReLU networks for square loss and orthogonal inputs

NEURIPS 2022arXiv

citations

Saddle-to-Saddle Dynamics in Diagonal Linear Networks

NEURIPS 2023arXiv

citations

Last iterate convergence of SGD for Least-Squares in the Interpolation regime.

NEURIPS 2021arXiv

citations

Sharpness-Aware Minimization Leads to Low-Rank Features

NEURIPS 2023arXiv

citations

Online Robust Regression via SGD on the l1 loss

NEURIPS 2020arXiv

citations

OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents

NEURIPS 2025arXiv

citations

Is In-Context Learning Sufficient for Instruction Following in LLMs?

ICLR 2025arXiv

citations

Penalising the biases in norm regularisation enforces sparsity

NEURIPS 2023arXiv

citations

Selective induction Heads: How Transformers Select Causal Structures in Context

ICLR 2025arXiv

citations

Learning In-context $n$-grams with Transformers: Sub-$n$-grams Are Near-Stationary Points

ICML 2025

citations

Transferable Adversarial Robustness for Categorical Data via Universal Robust Embeddings

NEURIPS 2023arXiv

citations

Long-Context Linear System Identification

ICLR 2025arXiv

citations

Continuized Accelerations of Deterministic and Stochastic Gradient Descents, and of Gossip Algorithms

NEURIPS 2021

citations

(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability

NEURIPS 2023

citations

On the spectral bias of two-layer linear networks

NEURIPS 2023

citations

Sequential Algorithms for Testing Closeness of Distributions

NEURIPS 2021

citations

Nicolas Flammarion

papers (21)

Square Attack: a query-efficient black-box adversarial attack via random search

Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks

Understanding and Improving Fast Adversarial Training

Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity

Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning

Gradient flow dynamics of shallow ReLU networks for square loss and orthogonal inputs

Saddle-to-Saddle Dynamics in Diagonal Linear Networks

Last iterate convergence of SGD for Least-Squares in the Interpolation regime.

Sharpness-Aware Minimization Leads to Low-Rank Features

Online Robust Regression via SGD on the l1 loss

OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents

Is In-Context Learning Sufficient for Instruction Following in LLMs?

Penalising the biases in norm regularisation enforces sparsity

Selective induction Heads: How Transformers Select Causal Structures in Context

Learning In-context $n$-grams with Transformers: Sub-$n$-grams Are Near-Stationary Points

Transferable Adversarial Robustness for Categorical Data via Universal Robust Embeddings

Long-Context Linear System Identification

Continuized Accelerations of Deterministic and Stochastic Gradient Descents, and of Gossip Algorithms

(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability

On the spectral bias of two-layer linear networks

Sequential Algorithms for Testing Closeness of Distributions

papers (21)

Square Attack: a query-efficient black-box adversarial attack via random search

Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks

Understanding and Improving Fast Adversarial Training

Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity

Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning

Gradient flow dynamics of shallow ReLU networks for square loss and orthogonal inputs

Saddle-to-Saddle Dynamics in Diagonal Linear Networks

Last iterate convergence of SGD for Least-Squares in the Interpolation regime.

Sharpness-Aware Minimization Leads to Low-Rank Features

Online Robust Regression via SGD on the l1 loss

OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents

Is In-Context Learning Sufficient for Instruction Following in LLMs?

Penalising the biases in norm regularisation enforces sparsity

Selective induction Heads: How Transformers Select Causal Structures in Context

Learning In-context $n$-grams with Transformers: Sub-$n$-grams Are Near-Stationary Points

Transferable Adversarial Robustness for Categorical Data via Universal Robust Embeddings

Long-Context Linear System Identification

Continuized Accelerations of Deterministic and Stochastic Gradient Descents, and of Gossip Algorithms

(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability

On the spectral bias of two-layer linear networks

Sequential Algorithms for Testing Closeness of Distributions