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.

Reasoning models are just LLMs

66 pointsby rognjen3 months ago

5 comments

iknownothow3 months ago
Is there a terminology battle happening in some circles? And if so, what are the consequences of being wrong and using the wrong terminology?<p>I follow the rnd and progress in this space and I haven&#x27;t heard anyone make a fuss about it. They are all LLMs or transformers or neural nets but they can be trained or optimized to do different things. For sure, there&#x27;s terms like Reasoning models or Chat models or Instruct models and yes they&#x27;re all LLMs.<p>But you can now start combining them to have hybrid models too. Are Omni models that handle audio and visual data still &quot;language&quot; models? This question is interesting in its own right for many reasons, but not to justify or bemoan the use of term LLM.<p>LLM is a good term, it&#x27;s a cultural term too. If you start getting pedantic, you&#x27;ll miss the bigger picture and possibly even the singularity ;)
评论 #43000922 未加载
评论 #43006800 未加载
评论 #43003977 未加载
janalsncm3 months ago
There was an HN thread that talked about how “just” is a 4 letter word. It significantly risks underestimating emergent properties and behaviors.<p>Every time you see “X is just Y” you should think of emergent behaviors. Complexity is difficult to predict.<p>&gt; R1 Zero has similar reasoning capabilities of R1 without requiring any SFT<p>In fact R1 zero was slightly better. This is an argument that RL and thinking tokens were a genuinely useful technique which I see as counter to the author’s thesis.<p>I also think a lot of what the author is referring to was more generously arguing against next token prediction (exact match of an answer) rather than the sequence-level rewards in R1.
jmor233 months ago
I think this is a response to Gary Marcus: <a href="https:&#x2F;&#x2F;xcancel.com&#x2F;GaryMarcus&#x2F;status&#x2F;1888606569245679893#m" rel="nofollow">https:&#x2F;&#x2F;xcancel.com&#x2F;GaryMarcus&#x2F;status&#x2F;1888606569245679893#m</a><p>“The architecture of the DeepSeek SYSTEM includes a model, and RL architecture that leverages symbolic rule.”<p>Marcus has long been a critic of deep learning and LLMs, saying they would “hit a wall”.
throwaway3141553 months ago
&gt; They say: “the progresses we are seeing are due to the fact that models like OpenAI o1 or DeepSeek R1 are not just LLMs”.<p>Would be nice if the author could cite even one example of this as it doesn&#x27;t match my experience whatsoever.
评论 #43005867 未加载
aaroninsf3 months ago
You and I are just networks of neurons (also). *<p>* applies only to meatreaders