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.

DeepSeek Laconic Decoding

1 pointsby svenfaw4 months ago

2 comments

eternityforest4 months ago
Can we weight the sampling based on a prediction of how long an answer will be?<p>Like with another model that just says &quot;Oh boy that word sounds like the beginning of an essay of nonsense, I don&#x27;t think that&#x27;s what I want to say&quot;?
svenfaw4 months ago
&quot;Discovered a very interesting thing about DeepSeek-R1 and all reasoning models: The wrong answers are much longer while the correct answers are much shorter.&quot;