Decoding-Time Language Model Alignment with Multiple Objectives

• URL: http://arxiv.org/abs/2406.18853v3
• Date: Mon, 28 Oct 2024 03:23:14 GMT
• Title: Decoding-Time Language Model Alignment with Multiple Objectives
• Authors: Ruizhe Shi, Yifang Chen, Yushi Hu, Alisa Liu, Hannaneh Hajishirzi, Noah A. Smith, Simon S. Du,
• Abstract summary: Existing methods primarily focus on optimizing LMs for a single reward function, limiting their adaptability to varied objectives. Here, we propose $textbfmulti-objective decoding (MOD)$, a decoding-time algorithm that outputs the next token from a linear combination of predictions. We show why existing approaches can be sub-optimal even in natural settings and obtain optimality guarantees for our method.
• Score: 116.42095026960598
• License:
• Abstract: Aligning language models (LMs) to human preferences has emerged as a critical pursuit, enabling these models to better serve diverse user needs. Existing methods primarily focus on optimizing LMs for a single reward function, limiting their adaptability to varied objectives. Here, we propose $\textbf{multi-objective decoding (MOD)}$, a decoding-time algorithm that outputs the next token from a linear combination of predictions of all base models, for any given weightings over different objectives. We exploit a common form among a family of $f$-divergence regularized alignment approaches (such as PPO, DPO, and their variants) to identify a closed-form solution by Legendre transform, and derive an efficient decoding strategy. Theoretically, we show why existing approaches can be sub-optimal even in natural settings and obtain optimality guarantees for our method. Empirical results demonstrate the effectiveness of the algorithm. For example, compared to a parameter-merging baseline, MOD achieves 12.8% overall reward improvement when equally optimizing towards $3$ objectives. Moreover, we experiment with MOD on combining three fully-finetuned LLMs of different model sizes, each aimed at different objectives such as safety, coding, and general user preference. Unlike traditional methods that require careful curation of a mixture of datasets to achieve comprehensive improvement, we can quickly experiment with preference weightings using MOD to find the best combination of models. Our best combination reduces toxicity on Toxigen to nearly 0% and achieves 7.9--33.3% improvement across other three metrics ($\textit{i.e.}$, Codex@1, GSM-COT, BBH-COT).

Related papers

• MM-RLHF: The Next Step Forward in Multimodal LLM Alignment [59.536850459059856]
  We introduce MM-RLHF, a dataset containing $mathbf120k$ fine-grained, human-annotated preference comparison pairs. We propose several key innovations to improve the quality of reward models and the efficiency of alignment algorithms. Our approach is rigorously evaluated across $mathbf10$ distinct dimensions and $mathbf27$ benchmarks.
  arXiv  Detail & Related papers  (2025-02-14T18:59:51Z)
• Test-time Alignment of Diffusion Models without Reward Over-optimization [8.981605934618349]
  Diffusion models excel in generative tasks, but aligning them with specific objectives remains challenging. We propose a training-free, test-time method based on Sequential Monte Carlo (SMC) to sample from the reward-aligned target distribution. We demonstrate its effectiveness in single-reward optimization, multi-objective scenarios, and online black-box optimization.
  arXiv  Detail & Related papers  (2025-01-10T09:10:30Z)
• Few-shot Steerable Alignment: Adapting Rewards and LLM Policies with Neural Processes [50.544186914115045]
  Large language models (LLMs) are increasingly embedded in everyday applications. Ensuring their alignment with the diverse preferences of individual users has become a critical challenge. We present a novel framework for few-shot steerable alignment.
  arXiv  Detail & Related papers  (2024-12-18T16:14:59Z)
• $f$-PO: Generalizing Preference Optimization with $f$-divergence Minimization [54.94545757220999]
  $f$-PO is a novel framework that generalizes and extends existing approaches. We conduct experiments on state-of-the-art language models using benchmark datasets.
  arXiv  Detail & Related papers  (2024-10-29T02:11:45Z)
• It's Morphing Time: Unleashing the Potential of Multiple LLMs via Multi-objective Optimization [16.54335356612006]
  The goal of model merging is to combine multiple models, each excelling in different tasks, into a single model that outperforms any of the individual source models. Existing methods rely heavily on human knowledge or intuition. It's difficult to obtain the great model merging configuration in limited evaluations.
  arXiv  Detail & Related papers  (2024-06-29T16:34:23Z)
• Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization [76.09576643028362]
  We present Multi-Objective Direct Preference Optimization (MODPO) for multiple alignment objectives. MODPO folds language modeling directly into reward modeling, training language models as implicit collective reward models. It theoretically yields the same optimal solutions as MORLHF but is practically more stable and efficient.
  arXiv  Detail & Related papers  (2023-10-05T17:35:26Z)
• Practical Bayesian Optimization of Objectives with Conditioning Variables [1.0497128347190048]
  We consider the more general case where a user is faced with multiple problems that each need to be optimized conditional on a state variable. Similarity across objectives boosts optimization of each objective in two ways. We propose a framework for conditional optimization: ConBO.
  arXiv  Detail & Related papers  (2020-02-23T22:06:26Z)
• Expected Information Maximization: Using the I-Projection for Mixture Density Estimation [22.096148237257644]
  Modelling highly multi-modal data is a challenging problem in machine learning. We present a new algorithm called Expected Information Maximization (EIM) for computing the I-projection. We show that our algorithm is much more effective in computing the I-projection than recent GAN approaches.
  arXiv  Detail & Related papers  (2020-01-23T17:24:50Z)
• AvgOut: A Simple Output-Probability Measure to Eliminate Dull Responses [97.50616524350123]
  We build dialogue models that are dynamically aware of what utterances or tokens are dull without any feature-engineering. The first model, MinAvgOut, directly maximizes the diversity score through the output distributions of each batch. The second model, Label Fine-Tuning (LFT), prepends to the source sequence a label continuously scaled by the diversity score to control the diversity level. The third model, RL, adopts Reinforcement Learning and treats the diversity score as a reward signal.
  arXiv  Detail & Related papers  (2020-01-15T18:32:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.