Instructive Decoding: Instruction-Tuned Large Language Models are Self-Refiner from Noisy Instructions
This addresses the limitation of instruction-tuned models in handling noisy or divergent instructions, offering a simple enhancement for better zero-shot generalization, though it is incremental as it builds on existing tuning methods.
The paper tackles the problem of instruction-tuned language models struggling with out-of-distribution instructions by proposing Instructive Decoding, which adjusts logits using noisy instructions to improve response accuracy, achieving considerable performance gains without parameter updates, with 'opposite' noisy instructions yielding the most significant improvements across models and tasks.
While instruction-tuned language models have demonstrated impressive zero-shot generalization, these models often struggle to generate accurate responses when faced with instructions that fall outside their training set. This paper presents Instructive Decoding (ID), a simple yet effective approach that augments the efficacy of instruction-tuned models. Specifically, ID adjusts the logits for next-token prediction in a contrastive manner, utilizing predictions generated from a manipulated version of the original instruction, referred to as a noisy instruction. This noisy instruction aims to elicit responses that could diverge from the intended instruction yet remain plausible. We conduct experiments across a spectrum of such noisy instructions, ranging from those that insert semantic noise via random words to others like 'opposite' that elicit the deviated responses. Our approach achieves considerable performance gains across various instruction-tuned models and tasks without necessitating any additional parameter updates. Notably, utilizing 'opposite' as the noisy instruction in ID, which exhibits the maximum divergence from the original instruction, consistently produces the most significant performance gains across multiple models and tasks.