Summary of The Power Of Active Multi-task Learning in Reinforcement Learning From Human Feedback, by Ruitao Chen et al.
The Power of Active Multi-Task Learning in Reinforcement Learning from Human Feedback
by Ruitao Chen, Liwei Wang
First submitted to arxiv on: 18 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 A reinforcement learning framework for large language models that leverages human feedback is explored. The approach involves multi-task representation learning, where a high-quality, low-dimensional representation is learned from a diverse set of source tasks. By formulating RLHF as a contextual dueling bandit problem and assuming a common linear representation, the sample complexity of source tasks can be reduced by considering task relevance. An algorithm is proposed to estimate task relevance using additional data and learn a policy. Theoretical guarantees are provided, showing that significant reductions in sample complexity can be achieved for both source and target tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Reinforcement learning from human feedback helps language models get better. To make this work with less human-labeled data, researchers use a technique called multi-task representation learning. They learn a simple way to represent many different tasks together. This paper shows how to make this process more efficient by considering which tasks are most important and allocating the right amount of data to each one. The authors also suggest an algorithm to figure out what makes certain tasks important, and prove that it can be very effective. |
Keywords
» Artificial intelligence » Multi task » Reinforcement learning » Reinforcement learning from human feedback » Representation learning » Rlhf
