Abstract:Aligning large language models (LLMs) to desirable human values requires balancing multiple, potentially conflicting objectives such as helpfulness, truthfulness, and harmlessness, which presents a multi-objective optimisation challenge. Most alignment pipelines rely on a fixed scalarisation of these objectives, which can introduce procedural unfairness by systematically under-weighting harder-to-optimise or minority objectives. To promote more equitable trade-offs, we introduce MGDA-Decoupled, a geometry-based multi-objective optimisation algorithm that finds a shared descent direction while explicitly accounting for each objective's convergence dynamics. In contrast to prior methods that depend on reinforcement learning (e.g., GAPO) or explicit reward models (e.g., MODPO), our approach operates entirely within the lightweight Direct Preference Optimisation (DPO) paradigm. Experiments on the UltraFeedback dataset show that geometry-aware methods -- and MGDA-Decoupled in particular -- achieve the highest win rates against golden responses, both overall and per objective.
| Comments: | Accepted to the Algorithmic Fairness Across Alignment Procedures and Agentic Systems Workshop at ICLR 2026 |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2604.20685 [cs.LG] |
| (or arXiv:2604.20685v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2604.20685 arXiv-issued DOI via DataCite (pending registration) |
From: Andor Vári-Kakas [view email]
[v1]
Wed, 22 Apr 2026 15:33:45 UTC (527 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。