Prompt and Parameter Co-Optimization for Large Language Models

2025-09-28

AI Summary

Existing research typically treats prompt optimization and parameter fine-tuning as disjoint paradigms, overlooking their synergistic potential. To address this, we propose MetaTuner—a unified framework that jointly models prompt learning and parameter adaptation. MetaTuner employs a dual-network architecture with a shared bottom-layer encoder to enable simultaneous explicit prompt generation and implicit parameter updating. To mitigate optimization challenges arising from coupling discrete prompt tokens with continuous model parameters, we introduce a supervised regularization loss. Extensive experiments across multiple benchmark tasks demonstrate that MetaTuner consistently outperforms pure prompt engineering, standard fine-tuning, and existing hybrid approaches. Results validate the effectiveness, robustness, and generalizability of co-optimizing prompts and parameters, establishing a new paradigm for efficient and adaptive language model adaptation.

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Search and Optimization: Metareasoning and Metaheuristics
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• 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

Abstract

Prompt optimization and fine-tuning are two major approaches to improve the performance of Large Language Models (LLMs). They enhance the capabilities of LLMs from complementary perspectives: the former through explicit natural language, and the latter through implicit parameter updates. However, prior work has typically studied them in isolation, leaving their synergistic potential largely underexplored. To bridge this gap, in this paper, we introduce MetaTuner, a novel framework that jointly integrates prompt optimization and fine-tuning for LLM training. Specifically, we introduce two neural networks to generate prompts and parameters, respectively, while allowing them to share a common bottom encoding layer to enable knowledge sharing. By the guidance of the final supervised signals, our framework is optimized to discover the optimal combinations between the prompts and parameters. Given that prompt learning involves discrete optimization while fine-tuning operates in a continuous parameter space, we design a supervised regularization loss to train our framework effectively. Extensive experiments across diverse benchmarks show that our method consistently outperforms the baselines.

Problem

• Jointly optimizing prompts and parameters for large language models
• Bridging isolated approaches of prompt optimization and fine-tuning
• Discovering optimal combinations between discrete prompts and continuous parameters

Innovation

• Jointly optimizes prompt generation and parameter fine-tuning
• Uses shared encoding layer for knowledge transfer
• Employs supervised regularization for discrete-continuous optimization