TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

ReLLM: Exact Structure for Large Language Model Completions

26 pointsby rckrdabout 2 years ago

5 comments

brian_cloutierabout 2 years ago
This is clever but seems to work through brute force and my guess is that results should be accompanied by some kind of confidence measure. You&#x27;re forcing the model to choose a path it wasn&#x27;t particularly interested in going down.<p>In the ideal case, such as when you ask for a json array, the first &quot;[&quot; token you enforce will have a relatively high probability and forcing it to go down that path will give you good results.<p>In the dangerous case the model doesn&#x27;t have a good structure-compliant completion to your prompt and the regex you supply forces the model into a path of extremely low probability and you get trash results.
评论 #35830457 未加载
评论 #35831255 未加载
Der_Einzigeabout 2 years ago
Is similar to this work on constrained text generation:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;hellisotherpeople&#x2F;constrained-text-generation-studio">https:&#x2F;&#x2F;github.com&#x2F;hellisotherpeople&#x2F;constrained-text-genera...</a>
评论 #35830755 未加载
killthebuddhaabout 2 years ago
This seems extremely similar to <a href="https:&#x2F;&#x2F;github.com&#x2F;newhouseb&#x2F;clownfish">https:&#x2F;&#x2F;github.com&#x2F;newhouseb&#x2F;clownfish</a>
评论 #35829987 未加载
_k1azabout 2 years ago
How’s this compared to ReLM: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2211.15458" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2211.15458</a>
fudged71about 2 years ago
My first thought was can you get the LLM to generate the pattern first, and then the completion?
评论 #35830508 未加载