Supplement Generation Training for Enhancing Agentic Task Performance

2026-04-22

AI Summary

This work addresses the challenges of high training costs, slow iteration cycles, and rapid obsolescence that hinder large language models (LLMs) in agent-based tasks, impeding their swift adaptation to new scenarios. To overcome these limitations, the authors propose Supplementary Generation Training (SGT), a lightweight approach that leverages a small auxiliary model to dynamically generate task-relevant supplementary text, which is then appended to the original input to enhance the LLM’s reasoning capabilities—without requiring any fine-tuning or parameter modification. By decoupling task-specific optimization from the base LLM, SGT enables low-cost, sustainable performance improvements. Experimental results demonstrate that SGT significantly boosts LLM performance across diverse agent tasks while substantially reducing training and deployment overhead, thereby supporting efficient and flexible real-world applications.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Training large foundation models for agentic tasks is increasingly impractical due to the high computational costs, long iteration cycles, and rapid obsolescence as new models are continuously released. Instead of post-training massive models for every new task or domain, we propose Supplement Generation Training (SGT), a more efficient and sustainable strategy. SGT trains a smaller LLM to generate useful supplemental text that, when appended to the original input, helps the larger LLM solve the task more effectively. These lightweight models can dynamically adapt supplements to task requirements, improving performance without modifying the underlying large models. This approach decouples task-specific optimization from large foundation models and enables more flexible, cost-effective deployment of LLM-powered agents in real-world applications.

Problem

• agentic tasks
• foundation models
• computational cost
• model obsolescence
• task performance

Innovation

• Supplement Generation Training
• agentic tasks
• large language models
• efficient adaptation
• modular AI