Summary of Prompt-based Length Controlled Generation with Reinforcement Learning, by Renlong Jie et al.
Prompt-Based Length Controlled Generation with Reinforcement Learning
by Renlong Jie, Xiaojun Meng, Lifeng Shang, Xin Jiang, Qun Liu
First submitted to arxiv on: 23 Aug 2023
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 This research proposes a prompt-based length control method for large language models (LLMs) like ChatGPT and GPT-4, enabling users to generate outputs of desired lengths. The autoregressive generation in LLMs is time-consuming, but controlling the generated length can reduce inference costs. To achieve high-accuracy length-controlled generation, the authors adopt reinforcement learning with a reward signal from trainable or rule-based reward models, rewarding outputs that follow pre-defined control instructions. They also introduce a standard prompt extractor to collect control information from users’ inputs. Experimental results show significant improvements in accuracy for summarization tasks on popular datasets like CNNDM and NYT. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study helps us understand how we can better use large language models to generate text of the right length. Imagine you want AI chatbots or language translation tools to give shorter or longer answers based on your needs. To make this happen, researchers created a new way to control the length of language model outputs using special prompts and rewards. They tested their method on two popular datasets and found it works well. This is important because it can help us get more useful results from AI language models in real-world scenarios. |
Keywords
» Artificial intelligence » Autoregressive » Gpt » Inference » Language model » Prompt » Reinforcement learning » Summarization » Translation
