Jugular
Jugular

Language Reasoning Models can overtake LLMs...

Here's my quick 3 minute breakdown:

  1. o1-preview: 97.8% on PlanBench Blocksworld vs. 62.5% for top LLMs, indicating shift from retrieval to reasoning.
  2. 52.8% on obfuscated "Mystery Blocksworld" vs. near-zero for LLMs, suggesting abstract reasoning skills, showing transfer capability.
  3. Variable "reasoning tokens" usage correlates with problem difficulty, hinting at internal search process, indicating adaptive compute.
1mo ago5.5K views
Jackietrader
Jackietrader

Adaptive compute is very interesting to me imo. I wonder how they are using variable compute for each task and basis what meta heuristic

StrangerDanger
StrangerDanger

Adaptive compute is what will help optimise cost for high complexity tasks, right?

Probably different "cores" for different types of tasks

SickleHammer
SickleHammer
Student1mo

Thanks for the paper! It's really interesting. I've been sounding like a madman explaining to people irl that Generative AI is not the end goal or even the natural next step of AI.

What is the next step?

BleakModem
BleakModem

Thanks for such a great post!!!! That's what I want more from this community.

Discover more
Curated from across
Kamekazi
KamekaziStudent20mo

What after Data analytics?

I am a fresher working as a Data analyst in a growing startup. My current role doesn't include much of coding just python and SQL and sometimes little bit ML work. Currently handling 2 projects, and some of PM work to define some roadmap...

Cremationsc
CremationscSelf Employed1mo

How LLMs are affecting coding interviews?

I know since the evolution of LLM tools like chat GPT ,sourcegraph. Many candidates are misusing them for coding interviews. There can be many hacks even in virtual interviews. Wouldn’t it results in shifting interview process to onsite ...

salt
saltGojek1mo

AI This Week: Liquid Foundation Models beat LLMs

Liquid AI has launched a new type of AI model called Liquid Foundation Models (LFMs). These models are designed to be more efficient than traditional transformer-based models, with smaller memory needs and better performance. LFMs come i...

Post image 1
Post image 2
11K views