Nicolas J-B. Brunel

Salim I. Amoukou, Nicolas J-B. Brunel

We analyze two widely used local attribution methods, Local Shapley Values and LIME, which aim to quantify the contribution of a feature value $x_i$ to a specific prediction $f(x_1, \dots, x_p)$. Despite their widespread use, we identify fundamental limitations in their ability to reliably detect locally important features, even under ideal conditions with exact computations and independent features. We argue that a sound local attribution method should not assign importance to features that neither influence the model output (e.g., features with zero coefficients in a linear model) nor exhibit statistical dependence with functionality-relevant features. We demonstrate that both Local SV and LIME violate this fundamental principle. To address this, we propose R-LOCO (Regional Leave Out COvariates), which bridges the gap between local and global explanations and provides more accurate attributions. R-LOCO segments the input space into regions with similar feature importance characteristics. It then applies global attribution methods within these regions, deriving an instance's feature contributions from its regional membership. This approach delivers more faithful local attributions while avoiding local explanation instability and preserving instance-specific detail often lost in global methods.

Salim I. Amoukou, Nicolas J-B. Brunel, Tangi Salaün

While Shapley Values (SV) are one of the gold standard for interpreting machine learning models, we show that they are still poorly understood, in particular in the presence of categorical variables or of variables of low importance. For instance, we show that the popular practice that consists in summing the SV of dummy variables is false as it provides wrong estimates of all the SV in the model and implies spurious interpretations. Based on the identification of null and active coalitions, and a coalitional version of the SV, we provide a correct computation and inference of important variables. Moreover, a Python library (All the experiments and simulations can be reproduced with the publicly available library Active Coalition of Variables, https://www.github.com/salimamoukou/acv00) that computes reliably conditional expectations and SV for tree-based models, is implemented and compared with state-of-the-art algorithms on toy models and real data sets.

Grégoire Martinon, Alexandra Lorenzo de Brionne, Jérôme Bohard et al.

The rise of generative AI, has driven significant advancements in high-risk sectors like healthcare and finance. The Retrieval-Augmented Generation (RAG) architecture, combining language models (LLMs) with search engines, is particularly notable for its ability to generate responses from document corpora. Despite its potential, the reliability of RAG systems in critical contexts remains a concern, with issues such as hallucinations persisting. This study evaluates a RAG system used in due diligence for an investment fund. We propose a robust evaluation protocol combining human annotations and LLM-Judge annotations to identify system failures, like hallucinations, off-topic, failed citations, and abstentions. Inspired by the Prediction Powered Inference (PPI) method, we achieve precise performance measurements with statistical guarantees. We provide a comprehensive dataset for further analysis. Our contributions aim to enhance the reliability and scalability of RAG systems evaluation protocols in industrial applications.

Vincent Blot, Alexandra Lorenzo de Brionne, Ines Sellami et al.

Image analysis is a key tool for describing the detailed mechanisms of folliculogenesis, such as evaluating the quantity of mouse Primordial ovarian Follicles (PMF) in the ovarian reserve. The development of high-resolution virtual slide scanners offers the possibility of quantifying, robustifying and accelerating the histopathological procedure. A major challenge for machine learning is to control the precision of predictions while enabling a high recall, in order to provide reproducibility. We use a multiple testing procedure that gives an overperforming way to solve the standard Precision-Recall trade-off that gives probabilistic guarantees on the precision. In addition, we significantly improve the overall performance of the models (increase of F1-score) by selecting the decision threshold using contextual biological information or using an auxiliary model. As it is model-agnostic, this contextual selection procedure paves the way to the development of a strategy that can improve the performance of any model without the need of retraining it.

Salim I. Amoukou, Nicolas J-B. Brunel, Tangi Salaün

Shapley Values (SV) are widely used in explainable AI, but their estimation and interpretation can be challenging, leading to inaccurate inferences and explanations. As a starting point, we remind an invariance principle for SV and derive the correct approach for computing the SV of categorical variables that are particularly sensitive to the encoding used. In the case of tree-based models, we introduce two estimators of Shapley Values that exploit the tree structure efficiently and are more accurate than state-of-the-art methods. Simulations and comparisons are performed with state-of-the-art algorithms and show the practical gain of our approach. Finally, we discuss the limitations of Shapley Values as a local explanation. These methods are available as a Python package.