Summary of Source2synth: Synthetic Data Generation and Curation Grounded in Real Data Sources, by Alisia Lupidi et al.
Source2Synth: Synthetic Data Generation and Curation Grounded in Real Data Sources
by Alisia Lupidi, Carlos Gemmell, Nicola Cancedda, Jane Dwivedi-Yu, Jason Weston, Jakob Foerster, Roberta Raileanu, Maria Lomeli
First submitted to arxiv on: 12 Sep 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 proposed Source2Synth method enables teaching large language models new skills without relying on human annotations. This approach takes a custom data source as input, generates synthetic data points with intermediate reasoning steps grounded in real-world sources, and improves dataset quality by discarding low-quality generations based on answerability. The technique is demonstrated in two challenging domains: multi-hop question answering (MHQA) and tabular question answering (TQA). Compared to fine-tuned baselines, Source2Synth achieves 25.51% improvement for TQA on WikiSQL and 22.57% improvement for MHQA on HotPotQA. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large Language Models struggle in scenarios that use structured data, complex reasoning, or tool usage. A new method called Source2Synth helps LLMs learn without needing expensive human help. It uses real-world sources to create new data points with steps that show how the answer was found. This makes the data better and more helpful. The approach is tested in two hard areas: asking questions that need multiple answers and using tools like tables to find answers. Source2Synth does much better than usual ways of doing this. |
Keywords
» Artificial intelligence » Question answering » Synthetic data
