Summary of Onegen: Efficient One-pass Unified Generation and Retrieval For Llms, by Jintian Zhang et al.
OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs
by Jintian Zhang, Cheng Peng, Mengshu Sun, Xiang Chen, Lei Liang, Zhiqiang Zhang, Jun Zhou, Huajun Chen, Ningyu Zhang
First submitted to arxiv on: 8 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Databases (cs.DB); Information Retrieval (cs.IR); 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 This paper presents a novel framework called OneGen that enables Large Language Models (LLMs) to simultaneously perform both generation and retrieval tasks in a single forward pass. The proposed approach incorporates autoregressively generated retrieval tokens, bridging the traditional training approaches for generation and retrieval. Experiments on composite tasks like RAG and Entity Linking demonstrate the effectiveness and efficiency of OneGen, showing improved retrieval performance while preserving the generative capabilities of LLMs. This unified framework, to the best of our knowledge, is the first to enable vector retrieval during generation in LLMs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research paper introduces a new way to improve computers’ ability to generate text and find information at the same time. The system, called OneGen, allows Large Language Models (LLMs) to do both tasks together, which is important for many real-world applications like searching and generating text. The authors tested their approach on two types of tasks and found that it works well and efficiently, while also keeping the original ability of LLMs to generate text. This breakthrough could lead to better AI systems in the future. |
Keywords
» Artificial intelligence » Entity linking » Rag
