Wow! I was exactly thinking to do Codenames I think it’s a really excellent fit for reasoning and world knowledge.
@karpathy
-
Games as Dynamic LLM Evaluation: Self-Balancing Difficulty Assessment
By
–
I quite like the idea using games to evaluate LLMs against each other, instead of fixed evals. Playing against another intelligent entity self-balances and adapts difficulty, so each eval (/environment) is leveraged a lot more. There's some early attempts around. Exciting area.
-
Teaching LLMs with Textbook Information Structure
By
–
We have to take the LLMs to school. When you open any textbook, you'll see three major types of information: 1. Background information / exposition. The meat of the textbook that explains concepts. As you attend over it, your brain is training on that data. This is equivalent
-
LLMs Need Education Like Humans Do
By
–
We just have to take the LLMs through school, exactly like humans.
-
TinyZero: Affordable RL Finetuning Under $30
By
–
TinyZero reproduction of R1-Zero
"experience the Ahah moment yourself for < $30" Given a base model, the RL finetuning can be relatively very cheap and quite accessible. -
Building Diverse RL Environments for LLM Cognitive Strategy Development
By
–
For friends of open source: imo the highest leverage thing you can do is help construct a high diversity of RL environments that help elicit LLM cognitive strategies. To build a gym of sorts. This is a highly parallelizable task, which favors a large community of collaborators.
-
Move 37: L’émergence de découvertes surprenantes en IA par renforcement
By
–
"Move 37" is the word-of-day – it's when an AI, trained via the trial-and-error process of reinforcement learning, discovers actions that are new, surprising, and secretly brilliant even to expert humans. It is a magical, just slightly unnerving, emergent phenomenon only
-
RL vs SL: Demystifying Reinforcement Learning’s Core Concept
By
–
Yeah exactly. I get triggered when RL is dressed up in its full rigorous math formalism because it's gate-keeping an essentially trivial core idea. SL:
a token sequence comes from some 3rd party source (e.g. human demonstration), and you just train on it. RL:
you first sample a -
Deep Learning’s Legendary Appetite for Compute Resources
By
–
I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent). I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed
-
Digital Agents as General-Purpose Tools Through Standard Interfaces
By
–
Projects like OpenAI’s Operator are to the digital world as Humanoid robots are to the physical world. One general setting (monitor keyboard and mouse, or human body) that can in principle gradually perform arbitrarily general tasks, via an I/O interface originally designed for… https://t.co/pZctgBa6PV
— Andrej Karpathy (@karpathy) 23 janvier 2025Projects like OpenAI’s Operator are to the digital world as Humanoid robots are to the physical world. One general setting (monitor keyboard and mouse, or human body) that can in principle gradually perform arbitrarily general tasks, via an I/O interface originally designed for