Summary of Seeing Eye to Ai: Human Alignment Via Gaze-based Response Rewards For Large Language Models, by Angela Lopez-cardona and Carlos Segura and Alexandros Karatzoglou and Sergi Abadal and Ioannis Arapakis
Seeing Eye to AI: Human Alignment via Gaze-Based Response Rewards for Large Language Models
by Angela Lopez-Cardona, Carlos Segura, Alexandros Karatzoglou, Sergi Abadal, Ioannis Arapakis
First submitted to arxiv on: 2 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
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 paper introduces GazeReward, a framework that integrates eye-tracking (ET) data into Reinforcement Learning from Human Feedback (RLHF) to improve the accuracy of Large Language Models (LLMs) in aligning their outputs with human expectations. The authors test different integration methods and models, demonstrating improved performance on established datasets. This work advances discussions on optimizing AI alignment with human values and explores the potential of cognitive data for future NLP research. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models like GPT, Llama, Claude, and Gemini are very good at doing many tasks, but they need to be trained a lot to match what humans think is correct. A way to do this is by using feedback from people, which helps the model learn what humans want it to say or do. The problem is that this method doesn’t always work well, because it’s hard to know exactly what humans want. In this paper, scientists created a new way to get feedback that uses information about where people are looking on their screens. This helps the model better understand what people prefer and makes its outputs more like what humans would say or do. |
Keywords
» Artificial intelligence » Alignment » Claude » Gemini » Gpt » Llama » Nlp » Reinforcement learning from human feedback » Rlhf » Tracking
