August 16, 2024 report
This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
preprint
trusted source
proofread
AI researchers introduce an LLM capable of generating text outputs of up to 10,000 words
![As existing LLMs fail to generate long enough output, AgentWrite adopts a plan-thenwrite pipeline to obtain a sufficient length output with off-the-shelf LLMs. Credit: arXiv (2024). DOI: 10.48550/arxiv.2408.07055 Chinese team introduces LongWriter—an LLM capable of generating text with more than 10,000 words](https://scx1.b-cdn.net/csz/news/800a/2024/chinese-team-introduce.jpg)
A team of AI researchers at Tsinghua University, working with a colleague from Zhipu AI, has developed a large language model (LLM) called LongWriter that they claim is capable of generating text output of up to 10,000 words. The group has written a paper describing their efforts and new LLM, which is available on the arXiv preprint server.
As LLMs have become mainstream, many have noticed that they are not capable of generating very long answers, such as full books or manuscripts—the current limit appears to be approximately 2,000 words. The researchers suggest this is because they are all trained on short documents. In their new effort, they have found that if LLMs are changed slightly and then trained using much longer documents, they are able to produce longer documents.
To test their idea, the research teams first trained a 9-billion parameter LLM using a conventional dataset, which included documents that were mostly less than 2,000 words long. As expected, when queried, it was not able to create texts longer than 2,000 words long.
Next, the team modified a traditional LLM using a pipeline they named AgentWrite to decompose training material into subtasks as it was processed. They then assembled a dataset they named "LongWriter-6k," which is a dataset that holds 6,000 written documents ranging in length from 2,000 to 32,000 words. They then trained the modified LLM using the new dataset LongWriter-6k and found that doing so increased the word length of documents it could produce to approximately 10,000 words.
In reviewing the newly produced long documents generated by the LLM, the team found them to be coherent and useable in a variety of contexts. They have posted the open-source code for their model on GitHub—a move that will allow others to build on what the team in China has done. They also posted a video showing LongWriter producing a 10,000-word tourist guide for people traveling in China.
The researchers acknowledge that there are ethical considerations that must be considered now that it has been found that LLMs can generate entire research papers, books, manuscripts or perhaps even movie scripts.
More information: Yushi Bai et al, LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs, arXiv (2024). DOI: 10.48550/arxiv.2408.07055
Github: github.com/THUDM/LongWriter
© 2024 Science X Network