RobuMTL: Enhancing Multi-Task Learning Robustness Against Weather Conditions

2026-01-16

AI Summary

This work addresses the significant performance degradation of multi-task learning models under adverse weather conditions, where visual degradation hinders model adaptability. To this end, the authors propose RobuMTL, a novel architecture that uniquely integrates dynamic Mixture-of-Experts (MoE) with hierarchical Low-Rank Adaptation (LoRA) to enable input-aware, task-specific adaptation. Specifically, RobuMTL dynamically selects task-dedicated LoRA modules and expert combinations based on the nature of input perturbations. This approach substantially enhances model robustness and generalization in complex weather scenarios, achieving an average relative improvement of +2.8% under single perturbations and up to +44.4% under mixed weather conditions on the PASCAL dataset, along with a cross-task average gain of +9.7% on NYUD-v2.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Robust Multi-Task Learning (MTL) is crucial for autonomous systems operating in real-world environments, where adverse weather conditions can severely degrade model performance and reliability. In this paper, we introduce RobuMTL, a novel architecture designed to adaptively address visual degradation by dynamically selecting task-specific hierarchical Low-Rank Adaptation (LoRA) modules and a LoRA expert squad based on input perturbations in a mixture-of-experts fashion. Our framework enables adaptive specialization based on input characteristics, improving robustness across diverse real-world conditions. To validate our approach, we evaluated it on the PASCAL and NYUD-v2 datasets and compared it against single-task models, standard MTL baselines, and state-of-the-art methods. On the PASCAL benchmark, RobuMTL delivers a +2.8% average relative improvement under single perturbations and up to +44.4% under mixed weather conditions compared to the MTL baseline. On NYUD-v2, RobuMTL achieves a +9.7% average relative improvement across tasks. The code is available at GitHub.

Problem

• Multi-Task Learning
• Robustness
• Weather Conditions
• Autonomous Systems
• Visual Degradation

Innovation

• Robust Multi-Task Learning
• Low-Rank Adaptation
• Mixture-of-Experts
• Weather Robustness
• Dynamic Module Selection