Summary of On Giant’s Shoulders: Effortless Weak to Strong by Dynamic Logits Fusion, By Chenghao Fan et al.
On Giant’s Shoulders: Effortless Weak to Strong by Dynamic Logits Fusion
by Chenghao Fan, Zhenyi Lu, Wei Wei, Jie Tian, Xiaoye Qu, Dangyang Chen, Yu Cheng
First submitted to arxiv on: 17 Jun 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 paper proposes a dynamic logit fusion approach to fine-tune large language models for task-specific applications without additional training. The method uses a series of task-specific small models, each specialized in a different task, and adaptively allocates weights among these models at each decoding step through Kullback-Leibler divergence constrained optimization problems. This approach achieves leading results across various benchmarks in both single-task and multi-task settings, closing the performance gap by 96.4% in single-task scenarios and by 86.3% in multi-task scenarios compared to full fine-tuning of the larger model. The method also surpasses performance on unseen tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about making big language models better at specific jobs without having to train them from scratch. It uses many smaller models, each good at a different task, and combines their strengths to create an even better model. This helps it learn faster and do better on new tasks. The results are really impressive, with the method doing much better than just training the big model alone. |
Keywords
* Artificial intelligence * Fine tuning * Multi task * Optimization
