Reasoning models are just LLMs
antirez.com
antirez.com
I follow the rnd and progress in this space and I haven'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's terms like Reasoning models or Chat models or Instruct models and yes they're all LLMs.
But you can now start combining them to have hybrid models too. Are Omni models that handle audio and visual data still "language" models? This question is interesting in its own right for many reasons, but not to justify or bemoan the use of term LLM.
LLM is a good term, it's a cultural term too. If you start getting pedantic, you'll miss the bigger picture and possibly even the singularity ;)
Alternatively I have seen debates as to what counts as a 'Small Language Model' that probably are nonsensical. Particularly because in my personal language war the term 'small language model' shouldn't even exist (no one knows that the threshold is, and our 'small' language models are bigger than the 'large' language models from just a few years ago).
This is fairly typical of new technology. Marketing departments will constantly come up with new terms or try to take over existing terms to push agendas. Terms with defined meaning will get abused by casual participants and loose all real meaning. Individuals new to the field will latch on to popular misuses of terms as they try to figure out what everyone is talking about and perpetuate definition creep. Old hands will overly focus on hair splitting exercises that no one else really cares about and sigh in dismay as their carefully cultured taxonomies collapse under expansion of interest in their field.
It will all work itself out in 10 years or so.
People need to know the performance they can expect from LLMs or agents. What are they capable of?
Result Specs (as in measuring output/experimental results) need strict definitions to be useful and I think the current ones with have for LLMs are pretty weak. (mostly benchmarks that model one kind of interaction, and usually not any sort of useful interaction)
To say a model is "just a LLM" is to presumably complain that it has no added bells or whistles that someone thinks is required beyond the above statistical model. And maybe I missed the point, but the author seems to be saying "yes it's just a LLM, but LLMs are all you need".
Every time you see “X is just Y” you should think of emergent behaviors. Complexity is difficult to predict.
> R1 Zero has similar reasoning capabilities of R1 without requiring any SFT
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.
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.
“The architecture of the DeepSeek SYSTEM includes a model, and RL architecture that leverages symbolic rule.”
Marcus has long been a critic of deep learning and LLMs, saying they would “hit a wall”.
Would be nice if the author could cite even one example of this as it doesn't match my experience whatsoever.
In any case I'll take your word for it, but that's still surprising to me.
Some argue they were kind of tricked into thinking that; see https://www.interconnects.ai/p/openais-o1-using-search-was-a... and some other writing by Lambert which has turned out pretty much on-point as far as RL and verifiers are concerned.
* applies only to meatreaders