arXiv subjects

Bruno Charron

Publications and source records attributed to Bruno Charron.

9 recordsLinked to original sources

arXiv2026

Interpretable Tabular Foundation Models via In-Context Kernel Regression

Tabular foundation models like TabPFN and TabICL achieve state-of-the-art performance through in-context learning, yet their architectures remain fundamentally opaque. We introduce KernelICL, a framework to enhance tabular foundation models with quantifiable sample-based interpretability. Building on the insight that in-context learning is akin to kernel regression, we make this mechanism explicit by replacing the final prediction layer with kernel functions (Gaussian, dot-product, kNN) so that every prediction is a transparent weighted average of training labels. We introduce a two-dimensional taxonomy that formally unifies standard kernel methods, modern neighbor-based approaches, and attention mechanisms under a single framework, and quantify inspectability via the perplexity of the weight distribution over training samples. On 55 TALENT benchmark datasets, KernelICL achieves performance on par with existing tabular foundation models, demonstrating that explicit kernel constraints on the final layer enable inspectable predictions without sacrificing performance.

cs.LG↗

arXiv2025

MEMERAG: A Multilingual End-to-End Meta-Evaluation Benchmark for Retrieval Augmented Generation

Automatic evaluation of retrieval augmented generation (RAG) systems relies on fine-grained dimensions like faithfulness and relevance, as judged by expert human annotators. Meta-evaluation benchmarks support the development of automatic evaluators that correlate well with human judgement. However, existing benchmarks predominantly focus on English or use translated data, which fails to capture cultural nuances. A native approach provides a better representation of the end user experience. In this work, we develop a Multilingual End-to-end Meta-Evaluation RAG benchmark (MEMERAG). Our benchmark builds on the popular MIRACL dataset, using native-language questions and generating responses with diverse large language models (LLMs), which are then assessed by expert annotators for faithfulness and relevance. We describe our annotation process and show that it achieves high inter-annotator agreement. We then analyse the performance of the answer-generating LLMs across languages as per the human evaluators. Finally we apply the dataset to our main use-case which is to benchmark multilingual automatic evaluators (LLM-as-a-judge). We show that our benchmark can reliably identify improvements offered by advanced prompting techniques and LLMs. Our dataset is available at https://github.com/amazon-science/MEMERAG

cs.CL↗

arXiv2021

PiRank: Scalable Learning To Rank via Differentiable Sorting

A key challenge with machine learning approaches for ranking is the gap between the performance metrics of interest and the surrogate loss functions that can be optimized with gradient-based methods. This gap arises because ranking metrics typically involve a sorting operation which is not differentiable w.r.t. the model parameters. Prior works have proposed surrogates that are loosely related to ranking metrics or simple smoothed versions thereof, and often fail to scale to real-world applications. We propose PiRank, a new class of differentiable surrogates for ranking, which employ a continuous, temperature-controlled relaxation to the sorting operator based on NeuralSort [1]. We show that PiRank exactly recovers the desired metrics in the limit of zero temperature and further propose a divide and-conquer extension that scales favorably to large list sizes, both in theory and practice. Empirically, we demonstrate the role of larger list sizes during training and show that PiRank significantly improves over comparable approaches on publicly available internet-scale learning-to-rank benchmarks.

cs.LG↗

arXiv2021

Large-scale Recommendation for Portfolio Optimization

Individual investors are now massively using online brokers to trade stocks with convenient interfaces and low fees, albeit losing the advice and personalization traditionally provided by full-service brokers. We frame the problem faced by online brokers of replicating this level of service in a low-cost and automated manner for a very large number of users. Because of the care required in recommending financial products, we focus on a risk-management approach tailored to each user's portfolio and risk profile. We show that our hybrid approach, based on Modern Portfolio Theory and Collaborative Filtering, provides a sound and effective solution. The method is applicable to stocks as well as other financial assets, and can be easily combined with various financial forecasting models. We validate our proposal by comparing it with several baselines in a domain expert-based study.

cs.AI↗

arXiv2015

Medium-heavy nuclei from nucleon-nucleon interactions in lattice QCD

On the basis of the Brueckner-Hartree-Fock method with the nucleon-nucleon forces obtained from lattice QCD simulations, the properties of the medium-heavy doubly-magic nuclei such as 16^O and 40^Ca are investigated. We found that those nuclei are bound for the pseudo-scalar meson mass M_PS ~ 470 MeV. The mass number dependence of the binding energies, single-particle spectra and density distributions are qualitatively consistent with those expected from empirical data at the physical point, although these hypothetical nuclei at heavy quark mass have smaller binding energies than the real nuclei.

hep-lat↗

arXiv2014

Charmed Tetraquarks Tcc and Tcs from Dynamical Lattice QCD Simulations

Charmed tetraquarks $T_{cc}=(cc\bar{u}\bar{d})$ and $T_{cs}=(cs\bar{u}\bar{d})$ are studied through the S-wave meson-meson interactions, $D$-$D$, $\bar{K}$-$D$, $D$-$D^{*}$ and $\bar{K}$-$D^{*}$, on the basis of the (2+1)-flavor lattice QCD simulations with the pion mass $m_π \simeq $410, 570 and 700 MeV. For the charm quark, the relativistic heavy quark action is employed to treat its dynamics on the lattice. Using the HAL QCD method, we extract the S-wave potentials in lattice QCD simulations, from which the meson-meson scattering phase shifts are calculated. The phase shifts in the isospin triplet ($I$=1) channels indicate repulsive interactions, while those in the $I=0$ channels suggest attraction, growing as $m_π$ decreases. This is particularly prominent in the $T_{cc} (J^P=1^+,I=0)$ channel, though neither bound state nor resonance are found in the range $m_π =410-700$ MeV. We make a qualitative comparison of our results with the phenomenological diquark picture.

hep-lat↗

arXiv2013

A comparative study of two lattice approaches to two-body systems

We present a method to extract the spectrum of two-particle systems on the lattice from wave functions computed in lattice simulations. The energies of the Hamiltonian eigenstates are extracted from the eigenvalues of a matrix, similar to a potential, constrained by the wave functions. This method is compared with the traditional variational method in the isospin 2 $ππ$ system.

hep-lat↗

arXiv2012

Pion-pion interaction in the I=1 channel

We present preliminary results of a new approach to the study of the pion-pion system in the I=1 channel. The Bethe-Salpeter wave function of the two-pion system is computed on the ground state and the first excited state. From these, we attempt to extract an interaction kernel (potential) which can then be used to extract observables such as the phase shifts. In a first trial, we use rather large pion masses $m_π\sim 1.05$ GeV and $m_π\sim 0.68$ GeV which do not allow rho decay.

hep-lat↗

arXiv2012

Construction of energy-independent potentials above inelastic thresholds in quantum field theories

We construct energy independent but non-local potentials above inelastic thresholds, in terms of Nambu-Bethe-Salpeter wave functions defined in quantum field theories such as QCD. As an explicit example, we consider NN --> NN + n pi scattering processes for n=0,1,2,.... We show an existence of energy-independent coupled channel potentials with a non-relativistic approximation, where momenta of all particles are small compared with their own masses. In the case of two-body inelastic scatterings such as Lambda Lambda --> Lambda Lambda, N Xi, Sigma Sigma, on the other hand, we show that energy-independent potentials can be constructed without relying on non-relativistic approximations. We also propose a method to extract these potentials using time-dependence of general correlation functions.

hep-lat↗