LORE: Lagrangian-Optimized Robust Embeddings for Visual Encoders

Proposed LORE, an unsupervised adversarial fine-tuning framework using constrained optimization to improve model robustness against attacks while preserving clean data accuracy. The method demonstrates enhanced zero-shot robustness, out-of-distribution generalization, and embedding interpretability with minimal performance degradation.

One Goal Many Challenges: Robust Preference Optimization Amid Content-Aware and Multi-Source Noise

Large Language Models (LLMs) have made significant strides in generating human-like responses, largely due to preference alignment techniques. However, these methods often assume unbiased human feedback, which is rarely the case in real-world scenarios.

Generalizable Spacecraft Trajectory Generation via Multimodal Learning with Transformers

Effective trajectory generation is essential for reliable on-board spacecraft autonomy. Among other approaches, learning-based warm-starting represents an appealing paradigm for solving the trajectory generation problem.

Cooperative Multi-Agent Constrained Stochastic Linear Bandits

We explore a collaborative multi-agent stochastic linear bandit setting involving a network of N agents that communicate locally to minimize their collective regret while keeping their expected cost under a specified threshold.