Continual Optimization with Symmetry Teleportation for Multi-Task Learning

2025-03-06

AI Summary

Gradient conflicts and task imbalance have long hindered performance in multi-task learning (MTL). This paper proposes a continual optimization framework grounded in symmetry transition: under loss-equivalence constraints, it dynamically identifies iso-loss alternative points in parameter space to mitigate gradient conflicts; it introduces a novel historical optimization trajectory reuse mechanism to enhance long-term convergence stability of optimizers; and it supports plug-and-play integration. The method synergistically combines low-rank adaptation (LoRA), loss-invariant objective design, gradient reweighting, and trajectory backtracking. Evaluated on multiple mainstream MTL benchmarks, it consistently surpasses state-of-the-art methods—both in absolute performance and robustness—while remaining fully compatible with and augmenting diverse advanced MTL approaches. The framework delivers stable, scalable, and cooperative multi-task optimization without architectural or training pipeline modifications.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Multi-task learning (MTL) is a widely explored paradigm that enables the simultaneous learning of multiple tasks using a single model. Despite numerous solutions, the key issues of optimization conflict and task imbalance remain under-addressed, limiting performance. Unlike existing optimization-based approaches that typically reweight task losses or gradients to mitigate conflicts or promote progress, we propose a novel approach based on Continual Optimization with Symmetry Teleportation (COST). During MTL optimization, when an optimization conflict arises, we seek an alternative loss-equivalent point on the loss landscape to reduce conflict. Specifically, we utilize a low-rank adapter (LoRA) to facilitate this practical teleportation by designing convergent, loss-invariant objectives. Additionally, we introduce a historical trajectory reuse strategy to continually leverage the benefits of advanced optimizers. Extensive experiments on multiple mainstream datasets demonstrate the effectiveness of our approach. COST is a plug-and-play solution that enhances a wide range of existing MTL methods. When integrated with state-of-the-art methods, COST achieves superior performance.

Problem

• Addresses optimization conflict in multi-task learning.
• Mitigates task imbalance using symmetry teleportation.
• Enhances MTL performance with loss-invariant objectives.

Innovation

• Uses symmetry teleportation for conflict reduction
• Implements low-rank adapters for loss-invariant objectives
• Applies historical trajectory reuse for optimizer benefits