Summary of Contrastive Instruction Tuning, by Tianyi Lorena Yan et al.
Contrastive Instruction Tuning
by Tianyi Lorena Yan, Fei Wang, James Y. Huang, Wenxuan Zhou, Fan Yin, Aram Galstyan, Wenpeng Yin, Muhao Chen
First submitted to arxiv on: 17 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Contrastive Instruction Tuning (CoIN) approach aims to improve the robustness and generalizability of large language models (LLMs) when faced with unseen instructions. This is achieved by maximizing the similarity between semantically equivalent instruction-instance pairs while minimizing the similarity between semantically different ones. To facilitate this approach, the authors augment the existing FLAN collection by paraphrasing task instructions. Experimental results on the PromptBench benchmark show that CoIN consistently improves LLMs’ accuracy when dealing with variations in character, word, sentence, and semantic levels, achieving an average increase of +2.5%. This technique has the potential to address trustworthiness issues in LLMs by improving their robustness to unseen instructions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models (LLMs) are super smart computers that can understand and generate human-like text. But they have a problem: they don’t do well when given instructions that are slightly different from what they’ve seen before. This means that if someone asks them to write about the same topic in a different way, they might not be able to. The authors of this paper want to fix this by making the LLMs more robust and able to understand different types of instructions. They came up with a new approach called Contrastive Instruction Tuning (CoIN) that helps the LLMs learn what makes different instructions similar or dissimilar. This way, when given an instruction they haven’t seen before, they can make better sense of it and generate more accurate text. |
Keywords
* Artificial intelligence * Instruction tuning
