Prompt-Level Distillation: A Non-Parametric Alternative to Model Fine-Tuning for Efficient Reasoning

TL;DR

Prompt-Level Distillation (PLD) offers a non-parametric, interpretable alternative to fine-tuning for efficient reasoning, significantly improving performance with minimal latency overhead and enabling human verification of logic.

Contribution

We introduce Prompt-Level Distillation, a method that extracts reasoning patterns into instructions for models, enhancing interpretability and efficiency without extensive fine-tuning.

Findings

Macro F1 scores improved from 57% to 90.0% and 67% to 83% on two datasets

Achieved frontier performance with negligible latency overhead

Enhanced interpretability for regulated industries

Abstract

Advanced reasoning typically requires Chain-of-Thought prompting, which is accurate but incurs prohibitive latency and substantial test-time inference costs. The standard alternative, fine-tuning smaller models, often sacrifices interpretability while introducing significant resource and operational overhead. To address these limitations, we introduce Prompt-Level Distillation (PLD). We extract explicit reasoning patterns from a Teacher model and organize them into a structured list of expressive instructions for the Student model's System Prompt. Evaluated on the StereoSet and Contract-NLI datasets using Gemma-3 4B, PLD improved Macro F1 scores from 57\% to 90.0\% and 67\% to 83\% respectively, enabling this compact model to match frontier performance with negligible latency overhead. These expressive instructions render the decision-making process transparent, allowing for full human verification of logic, making this approach ideal for regulated industries such as law, finance, and content moderation, as well as high-volume use cases and edge devices.