publications

2026

1. ICML 2026

   

   Tailoring Strictly Proper Scoring Rules for Downstream Tasks: An Application to Causal Inference

   Roman Plaud, Alexandre Perez-Lebel, Antoine Saillenfest, and 4 more authors

   2026

   Abs PDF Code

   Probabilistic models are typically trained using task-agnostic objectives (such as log-loss). This practice can lead to significant errors when these models are later used for downstream estimation tasks, a disconnect that is particularly critical in causal inference. This work investigates how to tailor strictly proper scoring rules to align better with specific downstream tasks, ensuring that the training objective is optimized for the intended application rather than just general probabilistic accuracy.

2025

1. ICML 2025

   

   To Each Metric Its Decoding: Post-Hoc Optimal Decision Rules of Probabilistic Hierarchical Classifiers

   Roman Plaud, Alexandre Perez-Lebel, Matthieu Labeau, and 2 more authors

   2025

   Abs PDF Code Poster

   Hierarchical classification offers an approach to incorporate the concept of mistake severity by leveraging a structured, labeled hierarchy. However, decoding in such settings frequently relies on heuristic decision rules, which may not align with task-specific evaluation metrics. In this work, we propose a framework for the optimal decoding of an output probability distribution with respect to a target metric. We derive optimal decision rules for increasingly complex prediction settings, providing universal algorithms when candidates are limited to the set of nodes. In the most general case of predicting a subset of nodes, we focus on rules dedicated to the hierarchical hFβ scores, tailored to hierarchical settings. To demonstrate the practical utility of our approach, we conduct extensive empirical evaluations, showcasing the superiority of our proposed optimal strategies, particularly in underdetermined scenarios. These results highlight the potential of our methods to enhance the performance and reliability of hierarchical classifiers in real-world applications.

2. NAACL

   

   Ignore the KL Penalty! Boosting Exploration on Critical Tokens to Enhance RL Fine-Tuning

   Jean Vassoyan, Nathanaël Beau, and Roman Plaud

   In Findings of the Association for Computational Linguistics: NAACL 2025 , Apr 2025

   Abs PDF Code

   The ability to achieve long-term goals is a key challenge in the current development of large language models (LLMs). To address this, pre-trained LLMs can be fine-tuned with reinforcement learning (RL) to explore solutions that optimize a given goal. However, exploration with LLMs is difficult, as a balance has to be struck between discovering new solutions and staying close enough to the pre-trained model, so as not to degrade basic capabilities. This is typically controlled with a Kullback-Leibler (KL) penalty. In this paper, we investigate the exploration dynamics of a small language model on a simple arithmetic task. We show how varying degrees of pre-training influence exploration and demonstrate the importance of “critical tokens” which have a dramatic impact on the final outcome. Consequently, we introduce a simple modification to the KL penalty that favors exploration on critical tokens, increasing the efficiency of the RL fine-tuning stage.

2024

1. CoNLL

   

   Revisiting Hierarchical Text Classification: Inference and Metrics

   Roman Plaud, Matthieu Labeau, Antoine Saillenfest, and 1 more author

   In Proceedings of the 28th Conference on Computational Natural Language Learning , Nov 2024

   Abs PDF Code Poster

   Hierarchical text classification (HTC) is the task of assigning labels to a text within a structured space organized as a hierarchy. Recent works treat HTC as a conventional multilabel classification problem, therefore evaluating it as such. We instead propose to evaluate models based on specifically designed hierarchical metrics and we demonstrate the intricacy of metric choice and prediction inference method. We introduce a new challenging dataset and we evaluate fairly, recent sophisticated models, comparing them with a range of simple but strong baselines, including a new theoretically motivated loss. Finally, we show that those baselines are very often competitive with the latest models. This highlights the importance of carefully considering the evaluation methodology when proposing new methods for HTC