Urgent.News

What's breaking now, across hundreds of outlets.

AI

Long-Context Generation Is a Sampling Problem (When You Turn Up the Heat)

Open-weight language models are widely observed to get worse the longer they generate. Often, attempts to fix this involve more training. At high temperature, The sampler, i.e. the rule that picks the next token out of the model's probabilities, decides whether long and agentic generation survives and/or thrives. We test 29 open-weight models from 1B to 122B parameters on ten established…

We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.

Read the original at arxiv.org →

Report a problem with this story

More in AI

Lessons from building a multilingual AI learning platform in Southeast Asia

Southeast Asia is one of the most multilingual regions in the world, and one of the most motivated to learn more languages. In Vietnam alone, students prepare for English exams, factory engineers study Japanese for jobs in Japan, and office workers take Chinese or Korean classes after hours because a new client or employer speaks […] The…

More from Tuesday 17 November →