Daily Digest — Sep 27

An Order-Sensitive Conflict Measure for Random Permutation Sets

• A novel order-sensitive conflict measure is introduced for Random Permutation Sets (RPSs) to address uncertainty and order information in fusion tasks. • The proposed measure analyzes conflicts from Dempster-Shafer Theory (DST) and Random Finite Set (RFS) perspectives, defining inconsistency as a graded, order-dependent concept.

Multiperspectivity as a Resource for Narrative Similarity Prediction

• An ensemble of 31 LLM personas, designed to capture multiperspectivity, achieves 0.705 accuracy in predicting narrative similarity on the SemEval-2026 Task 4 dataset. • Ensemble accuracy improves with size, showing Condorcet Jury Theorem-like dynamics, with practitioner personas yielding larger gains despite lower individual performance due to less correlated errors.

Online Fair Division with Additional Information

• In an online setting for fair division of indivisible goods, providing agents with normalization information (total values) enables algorithms to achieve stronger fairness guarantees than previously known.

Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data

• Discrete diffusion models with absorbing processes are shown to model the conditional distributions of clean data, offering a new analytical perspective. • Reparameterized Absorbing Discrete Diffusion (RADD) is introduced as a time-independent model that simplifies training and accelerates sampling by caching computations.