Pascal Vincent

Electric fields, forces, and modification of the tunnel barrier during field electron emission and field evaporation from single-wall carbon nanotubes

Vladimir Pimonov, Federico Panciera, Goulven Rouillé, Catherine Weng et autres

In this paper, we theoretically and experimentally investigate the specificities of single-walled carbon nanotubes (SWNTs) for field electron emission (FE) and field ion evaporation (FI). For FE, the small radii of curvature of these nanotubes lead to a significant widening of the …

fr, au (code pays fourni par la source)

Steering Large Language Model Activations in Sparse Spaces

Reza Bayat, Ali Rahimi-Kalahroudi, Mohammad Zakaria Pezeshki, Sarath Chandar et autres

A key challenge in AI alignment is guiding large language models (LLMs) to follow desired behaviors at test time. Activation steering, which modifies internal model activations during inference, offers a potential solution. However, prior work in dense activation spaces struggles with superposition, …

The Pitfalls of Memorization: When Memorization Hurts Generalization

Reza Bayat, Mohammad Pezeshki, Elvis Dohmatob, David López-Paz et autres

Neural networks often learn simple explanations that fit the majority of the data while memorizing exceptions that deviate from these explanations.This behavior leads to poor generalization when the learned explanations rely on spurious correlations. In this work, we formalize the interplay between …

Local thermal measurements and scanning thermal microscopy

Pierre‐Olivier Chapuis, Wenyu Zhao, Axel Pic, E. Guen et autres

National audience

fr, es, gb (code pays fourni par la source)

Compositional Risk Minimization

Divyat Mahajan, Mohammad Zakaria Pezeshki, Ioannis Mitliagkas, Kartik Ahuja et autres

Compositional generalization is a crucial step towards developing data-efficient intelligent machines that generalize in human-like ways. In this work, we tackle a challenging form of distribution shift, termed compositional shift, where some attribute combinations are completely absent at training but present in …

Investigation of Heat Transport in Individual Nanostructures by Scanning Thermal Microscopy

Séverine Gomès, Anselmo Acosta, G. Gadea-Diez, Felix R.M. Hamonangan M Hamonangan M et autres

International audience

fr, es (code pays fourni par la source)

Is the linear relationship between the slope and intercept observed in field emission S-K plots an artifact?

Anthony Ayari, Pascal Vincent, S. Perisanu, P. Poncharal et autres

The performance of field emitters is usually analyzed by linear fitting of a Fowler–Nordheim plot. It has sometimes been observed that the fitted slopes and intercepts show a strong correlation, but no convincing explanation has been provided. We propose a simple model …

fr (code pays fourni par la source)

Erratum: “All field emission experiments are noisy, … are any meaningful?” [J. Vac. Sci. Technol. B 41, 024001 (2023)]

Anthony Ayari, Pascal Vincent, S. Perisanu, P. Poncharal et autres

This is a correction to: All field emission experiments are noisy, … are any meaningful?

fr (code pays fourni par la source)

Field emission characterization of field-aligned carbon nanotubes synthesized in an environmental transmission electron microscope

Pascal Vincent, Federico Panciera, Ileana Florea, Anthony Ayari et autres

Optimizing the synthesis of carbon nanotubes (CNTs) for applications like field emission (FE) sources requires a fundamental understanding of the growth kinetics of individual CNTs. In this article, we explore how applying electric fields during CNT synthesis influences the as-grown nanotubes and …

fr, sg (code pays fourni par la source)

Local Heat Dissipation and Elasticity of Suspended Silicon Nanowires Revealed by Dual Scanning Electron and Thermal Microscopies

Jose Manuel Sojo Gordillo, Gerard Gadea, D. Renahy, Marc Salleras et autres

Abstract A novel combined setup, with a scanning thermal microscope (SThM) embedded in a scanning electron microscope (SEM), is used to characterize a suspended silicon rough nanowire (NW), which is epitaxially clamped at both sides and therefore monolithically integrated in a microfabricated …

ch, es, fr (code pays fourni par la source)

WorldSense: A Synthetic Benchmark for Grounded Reasoning in Large Language Models

Y. Benchekroun, Megi Dervishi, Mark Ibrahim, Jean-Baptiste Gaya et autres

We propose WorldSense, a benchmark designed to assess the extent to which LLMs are consistently able to sustain tacit world models, by testing how they draw simple inferences from descriptions of simple arrangements of entities. Worldsense is a synthetic benchmark with three …

Self-Supervised Disentanglement by Leveraging Structure in Data Augmentations

Cian Eastwood, Julius von Kügelgen, Linus Ericsson, Diane Bouchacourt et autres

Self-supervised representation learning often uses data augmentations to induce some invariance to "style" attributes of the data. However, with downstream tasks generally unknown at training time, it is difficult to deduce a priori which attributes of the data are indeed "style" and …