Summary of Importance Weighting Can Help Large Language Models Self-improve, by Chunyang Jiang et al.
Importance Weighting Can Help Large Language Models Self-Improve
by Chunyang Jiang, Chi-min Chan, Wei Xue, Qifeng Liu, Yike Guo
First submitted to arxiv on: 19 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 proposes a novel approach for large language model (LLM) self-improvement by filtering out correct but high distribution shift extent (DSE) samples. This method, called DS weight, approximates DSE using Importance Weighting methods and integrates it with self-consistency to comprehensively filter the self-generated samples and fine-tune the language model. The authors demonstrate that this approach can notably promote the reasoning ability of current LLM self-improvement methods, achieving performance comparable to those relying on external supervision from pre-trained reward models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about how computers can teach themselves to get better at tasks like answering questions and completing tasks. Right now, it’s hard and expensive to teach these computer programs using high-quality examples. So, researchers have been trying to figure out ways for the computer programs to learn from their own mistakes and improve themselves. One problem with this approach is that some of the examples they generate are bad or misleading. The paper introduces a new way to identify and get rid of these bad examples, which makes it easier for the computer program to learn and become better at its tasks. |
Keywords
* Artificial intelligence * Language model * Large language model
