LatentPrompt: Optimizing Promts in Latent Space

2025-08-04

AI Summary

Prompt engineering for large language models (LLMs) heavily relies on manual heuristics and lacks systematic, automated optimization. Method: This paper proposes the first model-agnostic prompt optimization framework operating in a continuous latent semantic space. It maps discrete prompts into embeddings and employs gradient-free, black-box optimization guided by automated evaluation metrics—requiring no manual rules or task-specific design. Initialization with only a seed prompt enables end-to-end automated prompt generation and iterative refinement. Contribution/Results: On financial sentiment classification, a single optimization round improves accuracy by approximately three percentage points, demonstrating effectiveness, generality, and plug-and-play usability. Crucially, this work pioneers the shift of prompt optimization from discrete token space to differentiable latent space, establishing a novel paradigm for automated prompt engineering.

Technology Category

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

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Recent advances have shown that optimizing prompts for Large Language Models (LLMs) can significantly improve task performance, yet many optimization techniques rely on heuristics or manual exploration. We present LatentPrompt, a model-agnostic framework for prompt optimization that leverages latent semantic space to automatically generate, evaluate, and refine candidate prompts without requiring hand-crafted rules. Beginning with a set of seed prompts, our method embeds them in a continuous latent space and systematically explores this space to identify prompts that maximize task-specific performance. In a proof-of-concept study on the Financial PhraseBank sentiment classification benchmark, LatentPrompt increased classification accuracy by approximately 3 percent after a single optimization cycle. The framework is broadly applicable, requiring only black-box access to an LLM and an automatic evaluation metric, making it suitable for diverse domains and tasks.

Problem

• Optimizing prompts in latent space for LLMs
• Automating prompt generation without manual rules
• Improving task performance via systematic prompt exploration

Innovation

• Optimizes prompts in latent semantic space
• Automatically generates and refines candidate prompts
• Requires only black-box LLM access