arXiv subjects

Thomas Bailleux

Publications and source records attributed to Thomas Bailleux.

4 recordsLinked to original sources

arXiv2026

Credal Concept Bottleneck Models for Epistemic-Aleatoric Uncertainty Decomposition

Concept Bottleneck Models (CBMs) predict through human-interpretable concepts, but they typically output point concept probabilities that conflate epistemic uncertainty (reducible model underspecification) with aleatoric uncertainty (irreducible input ambiguity). This makes concept-level uncertainty hard to interpret and, more importantly, hard to act upon. We introduce CREDENCE (Credal Ensemble Concept Estimation), a CBM framework that decomposes concept uncertainty by construction. CREDENCE represents each concept as a credal prediction (a probability interval), derives epistemic uncertainty from disagreement across diverse concept heads, and estimates aleatoric uncertainty via a dedicated ambiguity output trained to match annotator disagreement when available. The resulting signals support prescriptive decisions: automate low-uncertainty cases, prioritize data collection for high-epistemic cases, route high-aleatoric cases to human review, and abstain when both are high. Across several tasks, we show that epistemic uncertainty is positively associated with prediction errors, whereas aleatoric uncertainty closely tracks annotator disagreement, providing guidance beyond error correlation. Our implementation is available at the following link: https://github.com/Tankiit/Credal_Sets/tree/ensemble-credal-cbm

cs.AI↗

arXiv2026

Explanation Quality Assessment as Ranking with Listwise Rewards

We reformulate explanation quality assessment as a ranking problem rather than a generation problem. Instead of optimizing models to produce a single "best" explanation token-by-token, we train reward models to discriminate among multiple candidate explanations and learn their relative quality. Concretely, we construct per-instance candidate sets with graded quality levels and train listwise and pairwise ranking models (ListNet, LambdaRank, RankNet) to preserve ordinal structure and avoid score compression typical of pointwise regression or binary preference objectives. We observe three findings: First, ranking losses consistently outperform regression on score separation across all domains tested. Second, the optimal ranking loss depends on data characteristics: listwise objectives excel with well-separated quality tiers, while pairwise methods are more robust to noisy natural annotations. Third, when trained on carefully curated and well-structured data, small encoder models can match models that are orders of magnitude larger, suggesting that data quality matters more than model scale. Finally, when used as rewards in policy optimization, ranking-based scores enable stable convergence in settings where regression-based rewards fail entirely. Code and data are available at: https://github.com/Tankiit/PPO_Learning_to_rank

cs.AI↗

arXiv2024

TurboTLS: TLS connection establishment with 1 less round trip

We show how to establish TLS connections using one less round trip. In our approach, which we call TurboTLS, the initial client-to-server and server-to-client flows of the TLS handshake are sent over UDP rather than TCP. At the same time, in the same flights, the three-way TCP handshake is carried out. Once the TCP connection is established, the client and server can complete the final flight of the TLS handshake over the TCP connection and continue using it for application data. No changes are made to the contents of the TLS handshake protocol, only its delivery mechanism. We avoid problems with UDP fragmentation by using request-based fragmentation, in which the client sends in advance enough UDP requests to provide sufficient room for the server to fit its response with one response packet per request packet. Clients can detect which servers support this without an additional round trip, if the server advertises its support in a DNS HTTPS resource record. Experiments using our software implementation show substantial latency improvements. On reliable connections, we effectively eliminate a round trip without any noticeable cost. To ensure adequate performance on unreliable connections, we use lightweight packet ordering and buffering; we can have a client wait a very small time to receive a potentially lost packet (e.g., a fraction of the RTT observed for the first fragment) before falling back to TCP without any further delay, since the TCP connection was already in the process of being established. This approach offers substantial performance improvements with low complexity, even in heterogeneous network environments with poorly configured middleboxes.

cs.CR↗

arXiv2024

Ontology Completion with Natural Language Inference and Concept Embeddings: An Analysis

We consider the problem of finding plausible knowledge that is missing from a given ontology, as a generalisation of the well-studied taxonomy expansion task. One line of work treats this task as a Natural Language Inference (NLI) problem, thus relying on the knowledge captured by language models to identify the missing knowledge. Another line of work uses concept embeddings to identify what different concepts have in common, taking inspiration from cognitive models for category based induction. These two approaches are intuitively complementary, but their effectiveness has not yet been compared. In this paper, we introduce a benchmark for evaluating ontology completion methods and thoroughly analyse the strengths and weaknesses of both approaches. We find that both approaches are indeed complementary, with hybrid strategies achieving the best overall results. We also find that the task is highly challenging for Large Language Models, even after fine-tuning.

cs.CL↗