Steer Model beyond Assistant: Controlling System Prompt Strength via Contrastive Decoding

TL;DR

This paper presents a training-free method called system prompt strength that uses contrastive decoding to dynamically control large language model behavior, improving adherence to specific system prompts across various tasks.

Contribution

It introduces a novel contrastive decoding technique to modulate system prompt influence without retraining, enhancing control over model behavior.

Findings

Up to +8.5 accuracy on IFEval

+45pp refusal rate on OffTopicEval

+13% steerability on Prompt-Steering

Abstract

Large language models excel at complex instructions yet struggle to deviate from their helpful assistant persona, as post-training instills strong priors that resist conflicting instructions. We introduce system prompt strength, a training-free method that treats prompt adherence as a continuous control. By contrasting logits from target and default system prompts, we isolate and amplify the behavioral signal unique to the target persona by a scalar factor alpha. Across five diverse benchmarks spanning constraint satisfaction, behavioral control, pluralistic alignment, capability modulation, and stylistic control, our method yields substantial improvements: up to +8.5 strict accuracy on IFEval, +45pp refusal rate on OffTopicEval, and +13% steerability on Prompt-Steering. Our approach enables practitioners to modulate system prompt strength, providing dynamic control over model behavior without retraining.