INRIA a CCSD electronic archive server
Not a member yet
122212 research outputs found
Sort by
MObyGaze: a film dataset of multimodal objectification densely annotated by experts
No full textCharacterizing and quantifying gender representation disparities in audiovisual storytelling contents is necessary to grasp how stereotypes may perpetuate on screen. In this article, we consider the high-level construct of objectification and introduce a new AI task to the ML community: characterize and quantify complex multimodal (visual, speech, audio) temporal patterns producing objectification in films. Building on film studies and psychology, we define the construct of objectification in a structured thesaurus involving 5 sub-constructs manifesting through 11 concepts spanning 3 modalities. We introduce the Multimodal Objectifying Gaze (MObyGaze) dataset, made of 20 movies annotated densely by experts for objectification levels and concepts over freely delimited segments: it amounts to 6072 segments over 43 hours of video with fine-grained localization and categorization. We formulate different learning tasks, propose and investigate best ways to learn from the diversity of labels among a low number of annotators, and benchmark recent vision, text and audio models, showing the feasibility of the task. We make our code and our dataset available to the community and described in the Croissant format: https://anonymous.4open.science/r/MObyGaze-F600/ Related worksWe position our contributions with respect to works on: analyses of biases in film datasets, annotation of audiovisual and mulimodal contents, and dataset creation for interpretive tasks.</div
A Note on k-NN Gating in RAG
No full textWe develop a statistical proxy framework for retrieval-augmented generation (RAG), designedto formalize how a language model (LM) should balance its own predictions with retrievedevidence. For each query x, the system combines a frozen base model q0 (· x) with a k-nearestneighbor retriever r (k ) (· x) through a measurable gate k(x). A retrieval-trust weight wfact (x)quantifies the geometric reliability of the retrieved neighborhood and penalizes retrieval in low-trust regions. We derive the Bayes-optimal per-query gate and analyze its effect on a discordance-based hallucination criterion that captures disagreements between LM predictions and retrievedevidence. We further show that this discordance admits a deterministic asymptotic limit governedsolely by the structural agreement (or disagreement) between the Bayes rule and the LM. Toaccount for distribution mismatch between queries and memory, we introduce a hybrid geometric-semantic model combining covariate deformation and label corruption. Overall, this note providesa principled statistical foundation for factuality-oriented RAG systems
Motion Artifact Removal from EEG Signals Using the Motion-Net Deep Learning Algorithm
No full textInternational audienceElectroencephalography (EEG) is vital in brain signal analysis, with mobile EEG (mo-EEG) extending its utility to movement scenarios. However, mo-EEG is highly susceptible to motion artifacts, degrading signal quality. This study introduces Motion-Net, a subject-specific CNN-based framework for removing motion artifacts from EEG signals. Unlike prior generalized or simulated approaches, Motion-Net is trained and tested separately per subject using real EEG recordings with ground-truth references. The model incorporates visibility graph (VG) features, providing structural information that improves performance with smaller datasets. Across three experimental setups, Motion-Net achieved an average motion artifact reduction percentage () of 86% ±4.13, a SNR improvement of 20 ±4.47 dB, and a Mean Absolute Error (MAE) of 0.20 ±0.16. These results show that VG features enhance model learning and stability, and establish Motion-Net as the first subject-specific deep learning framework for robust motion artifact removal in mobile EEG
Exact Minimum Cuts in Hypergraphs at Scale
No full textInternational audienceThe hypergraph minimum cut problem aims to partition the vertices of a hypergraph into two non-empty parts while minimizing the total weight of hyperedges crossing the cut. This problem lies at the core of many tasks in network reliability, VLSI placement, and community detection. We introduce HeiCut, the first algorithm that makes exact minimum cut computation feasible for both weighted and unweighted instances at scales of hundreds of millions of vertices. HeiCut presents seven exact reduction rules that provably preserve the minimum cut, and an optional heuristic contraction based on label propagation that shrinks complex and persistent structures. When no further reductions are possible, the remaining in stance is solved exactly with a known algorithm. Our extensive evaluation on more than 500 real-world hypergraphs reveals that the exact reductions alone already expose the minimum cut (i.e., the residual collapses to a single vertex or has no hyperedges) in over 85% of instances. Across all instances, HeiCut solves over twice as many instances as the state-of-the-art within set computational limits, and is up to five orders of magnitude faster. Thus, HeiCut significantly advances hyper graph minimum cut computation in real-world, large-scale scenarios
CoPubli: a minimal software pipeline to extract, browse and geolocalise HAL coauthors
No full textThis publication describes CoPubli, a software created to retrieve Inria publications involving foreign co-authors, in order to help develop international cooperation between research teams and authors from institutions around the world. The software uses the XML-TEI rendering of the HAL open archive API and analyses, with Python Jupyter Notebook, the contents to determine the city of each stated institution. A Dashboard offers different views of the results, such as a map, a keyword cloud, four year evolution of copublications. The views can be refined combining different filters (Research team, foreign institution, city, country, etc.)
Forward self-similar solutions to the 2D Navier--Stokes equations
No full textWe construct self-similar solutions to the 2D Navier--Stokes equations evolving from arbitrarily large --homogeneous initial data and present numerical evidence for their non-uniqueness
Cascading predictions from common to uncommon species improves species distribution models for plants
No full textInternational audienceSpecies distribution models (SDMs) traditionally rely on abiotic factors like climate and topography to predict plant species distributions. While effective at broad scales, these models often fail at finer spatial resolutions due to their inability to capture localized environmental conditions and biotic interactions, such as competition and facilitation, that strongly influence species presence. To address these limitations, we propose a cascading prediction framework that leverages species co-occurrence relationships to improve SDM predictions especially at small spatial scales. In this approach, we first predict the presence of common, dominant species based on environmental data and then use these predictions to inform the presence of less common species. We explore two variations: (i) the Predictive Cascade, which uses model-based predictions of frequent species to help predict the remaining species, and (ii) the Disjunctive Observational Cascade, which integrates presence-only data from citizen science platforms to the Cascade pipeline. By incorporating biotic interactions and competitive hierarchies into SDMs, our cascading approach constitutes a novel method to enhance prediction accuracy at fine spatial resolutions, particularly in species-rich environments where current state-of-the-art models struggle
CausalProfiler: Generating Synthetic Benchmarks for Rigorous and Transparent Evaluation of Causal Machine Learning
No full textInternational audienceCausal machine learning (Causal ML) aims to answer "what if" questions using machine learning algorithms, making it a promising tool for high-stakes decision-making. Yet, empirical evaluation practices in Causal ML remain limited. Existing benchmarks often rely on a handful of hand-crafted or semi-synthetic datasets, leading to brittle, non-generalizable conclusions. To bridge this gap, we introduce CausalProfiler, a synthetic benchmark generator for Causal ML methods. Based on a set of explicit design choices about the class of causal models, queries, and data considered, the CausalProfiler randomly samples causal models, data, queries, and ground truths constituting the synthetic causal benchmarks. In this way, Causal ML methods can be rigorously and transparently evaluated under a variety of conditions. This work offers the first random generator of synthetic causal benchmarks with coverage guarantees and transparent assumptions operating on the three levels of causal reasoning: observation, intervention, and counterfactual. We demonstrate its utility by evaluating several state-of-the-art methods under diverse conditions and assumptions, both in and out of the identification regime, illustrating the types of analyses and insights the CausalProfiler enables
On optimal transport with f-divergence regularization
No full textThis report establishes that an optimal transport (OT) problem regularized by a given f-divergence admits the same solution as another OT problem regularized by a different gdivergence, under an appropriate transformation of the cost function. This structural equivalence between OT problems regularized by distinct divergences, in the sense of sharing the same unique minimizer, is demonstrated within the framework of polish spaces with bounded cost functions. Under these general assumptions, the existence of optimal potentials for the dual problem and the uniqueness of the optimal coupling for the primal problem are rigorously established
The Generation Phases of Flow Matching: a Denoising Perspective
No full textFlow matching has achieved remarkable success, yet the factors influencing the quality of its generation process remain poorly understood. In this work, we adopt a denoising perspective and design a framework to empirically probe the generation process. Laying down the formal connections between flow matching models and denoisers, we provide a common ground to compare their performances on generation and denoising. This enables the design of principled and controlled perturbations to influence sample generation: noise and drift. This leads to new insights on the distinct dynamical phases of the generative process, enabling us to precisely characterize at which stage of the generative process denoisers succeed or fail and why this matters