Summary of Hammer: Robust Function-calling For On-device Language Models Via Function Masking, by Qiqiang Lin et al.
Hammer: Robust Function-Calling for On-Device Language Models via Function Masking
by Qiqiang Lin, Muning Wen, Qiuying Peng, Guanyu Nie, Junwei Liao, Jun Wang, Xiaoyun Mo, Jiamu Zhou, Cheng Cheng, Yin Zhao, Jun Wang, Weinan Zhang
First submitted to arxiv on: 6 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
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 Large language models have shown great potential when equipped with external tools and API calls. However, to truly unlock their capabilities for complex tasks, enhancements in their function calling abilities are crucial. This paper highlights a significant gap in existing function calling models, where performance varies greatly across benchmarks due to misleading naming conventions. To address this issue, the authors introduce Hammer, a family of foundation models engineered for on-device function calling. Hammer employs an augmented dataset and function masking techniques to enhance models’ sensitivity to irrelevant functions. The results show that Hammer outperforms larger models and achieves state-of-the-art (SOTA) results across diverse benchmarks. This paper’s open-source contributions include a specialized dataset, a tuning framework, and the Hammer models, setting a new standard for function calling performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models are very good at doing things on their own when given special tools. But to really use them well, we need to improve how they find and use functions. This paper talks about a big problem with existing ways of finding functions – sometimes they get confused by silly names! To fix this, the authors created something called Hammer, which is special because it can find functions better than other models. They tested Hammer on lots of different tasks and it did really well – better than bigger models! The authors are sharing their work so that others can use it too. |
