Instruction Fine-Tuning, a method enhancing pre-trained language models’ capabilities from mere next-word prediction to complex instruction following, often employs a one-off training approach on diverse instruction dataset. However, this method may not effectively enhance models’ adherence to instructions due to the simultaneous handling of varying instruction complexities. To address this, we propose a novel phased instruction fine-tuning (Phased IFT) method, grounded in the hypothesis of progressive alignment, which posits that the transition of a pre-trained language model from simple next-word prediction to sophisticated instruction following is a gradual learning process. Specifically, we obtain the score of difficulty for each instruction via GPT-4, stratify the instruction data into subsets of increasing difficulty, and sequentially uptrain on these subsets using the standard supervised loss. Through extensive experiments on the pre-trained models Llama-2 7B/13B, and Mistral-7B using the 52K Alpaca instruction data, we demonstrate that Phased IFT significantly surpasses traditional one-off instruction fine-tuning (One-off IFT) method in win rate, empirically validating the progressive alignment hypothesis. Our findings suggest that Phased IFT offers a simple yet effective pathway for elevating the instruction-following capabilities of pre-trained language models.

Robots are often deployed in remote locations for tasks such as exploration, where users cannot directly perceive the agent and its environment. For Human-In-The-Loop applications, operators must have a comprehensive understanding of the robot’s current state and its environment to take necessary actions and effectively assist the agent. In this work, we compare different explanation styles to determine the most effective way to convey real-time updates to users. Additionally, we formulate these explanation styles as separate fine-tuning tasks and assess the effectiveness of large language models in delivering in-mission updates to maintain situation awareness. The code and dataset for this work are available at:———

This paper presents a study to understand the issues related to using NLG to humanise explanations from a popular interpretable machine learning framework called LIME. Our study shows that self-reported rating of NLG explanation was higher than that for a non-NLG explanation. However, when tested for comprehension, the results were not as clear-cut showing the need for performing more studies to uncover the factors responsible for high-quality NLG explanations.