It's against HN's guidelines to editorialize titles like this. From <a href="https://news.ycombinator.com/newsguidelines.html">https://news.ycombinator.com/newsguidelines.html</a>:<p>"<i>Please use the original title, unless it is misleading or linkbait; don't editorialize.</i>"<p>If you want to say what you think is important about an article, that's fine, but do it by adding a comment to the thread. Then your view will be on a level playing field with everyone else's: <a href="https://hn.algolia.com/?dateRange=all&page=0&prefix=false&sort=byDate&type=comment&query=%22level%20playing%20field%22%20by:dang" rel="nofollow noreferrer">https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...</a>
I just want the 100k context window Anthropic gave Claude, but for GPT. Claude likes to hallucinate when I ask it to build chapter notes for a class I’m taking, and I don’t want to have to break up the text into tiny bits for GPT…
Link to the paper directly <a href="https://arxiv.org/pdf/2310.17680.pdf" rel="nofollow noreferrer">https://arxiv.org/pdf/2310.17680.pdf</a><p>The system that they describe, Codefusion is interesting because it's a diffusion model for generating code rather than an autoregressive model like most LLM code generators.
This was expected because open source models of the same size already beat GPT-3.5 in many ways. And Mistral 7B makes you think if huge parameters are even needed for something like GPT-3.5 level.
I could be wrong but I believe the original comment that's used as the title here comes from this tweet: <a href="https://twitter.com/felix_red_panda/status/1718916631512949248?t=jmCkeVH1Hyyu4vmwY-NhQg&s=19" rel="nofollow noreferrer">https://twitter.com/felix_red_panda/status/17189166315129492...</a>