# Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction Source: [Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction]() Collection index: [llms.txt]() Data updated: 2026-05-16T05:00:12.406Z Authors: Yuanzhe Li, Jianing Deng, Jingtong Hu, Tianlong Chen, Song Wang, Huanrui Yang Published: 2026-02-02T19:24:04.000Z ## Abstract Large language models (LLMs) achieve strong performance in long-horizon decision-making tasks through multi-step interaction and reasoning at test time. While practitioners commonly believe a higher task success rate necessitates the use of a larger and stronger LLM model, multi-step interaction with a large LLM incurs prohibitive inference cost. To address this problem, we explore the use of low-precision quantized LLMs in the long-horizon decision-making process. Based on the observation of diverse sensitivities among interaction steps, we propose Dynamic Mixed-Precision Routing (DMR), a framework that adaptively selects between high-precision and low-precision LLMs at each decision step. The router is trained via a two-stage pipeline, consisting of KL-divergence-based supervised learning that identifies precision-sensitive steps, followed by Group-Relative Policy Optimization (GRPO) to further improve task success rates. Experiments on ALFWorld and WebShop demonstrate that our approach achieves a strong accuracy-cost trade-off over single-precision baselines. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Multi-step LLM Interaction on Qwen3-8B (test)](): Multi-step LLM Interaction; Qwen3-8B Inference Performance (test); High-Precision Ratio: 26.7. - [Multi-step LLM Interaction on Qwen3-1.7B Inference Performance (test)](): Multi-step LLM Interaction; Qwen3-1.7B Inference Performance (test); High-Precision Ratio: 20.5. - [Multi-step LLM Interaction on Qwen3-4B Inference Performance (test)](): Multi-step LLM Interaction; Qwen3-4B Inference Performance (test); High-Precision Ratio: 8.6. - [Multi-step LLM Interaction on DeepSeek-R1-Distill-Llama-8B Inference Performance (test)](): Multi-step LLM Interaction; DeepSeek-R1-Distill-Llama-8B Inference Performance (test); High-Precision Ratio: 9.8. ## Original sources and implementation [Original paper]() [PDF]()