Contrastive Instruction Tuning

TL;DR

Contrastive Instruction Tuning enhances large language models' robustness to unseen instructions by aligning semantically equivalent instruction representations, leading to more consistent outputs across varied instruction phrasings.

Contribution

This paper introduces a novel contrastive training method for instruction tuning that improves LLM robustness to instruction variations by augmenting instruction data with paraphrases.

Findings

Improves robustness to unseen instructions by +2.5% accuracy on PromptBench

Enhances consistency of LLM outputs across instruction variations

Demonstrates effectiveness across multiple levels of textual variation

Abstract

Instruction tuning has been used as a promising approach to improve the performance of large language models (LLMs) on unseen tasks. However, current LLMs exhibit limited robustness to unseen instructions, generating inconsistent outputs when the same instruction is phrased with slightly varied forms or language styles. This behavior indicates LLMs' lack of robustness to textual variations and generalizability to unseen instructions, potentially leading to trustworthiness issues. Accordingly, we propose Contrastive Instruction Tuning, which maximizes the similarity between the hidden representations of semantically equivalent instruction-instance pairs while minimizing the similarity between semantically different ones. To facilitate this approach, we augment the existing FLAN collection by paraphrasing task instructions. Experiments on the PromptBench benchmark show that CoIN consistently improves LLMs' robustness to unseen instructions with variations across character, word, sentence, and semantic levels by an average of +2.5% in accuracy. Code is available at https://github.com/luka-group/CoIN.