Process supervision by an LLM judge cannot be run for long, because a judge with billions of parameters is gameable and RL will find its cracks.

Added to korrents 7 Sept 2026 · How quotes work · Something wrong? Tell us

Shared subject first, then similar wording. They may agree or disagree.

• Large language models mimic what people say to do rather than work out what to do, which is why they are not about understanding the world. RSRichard Sutton26 Sept 2025LLMsreinforcement learning
• Reinforcement learning cannot be scaled for robots the way it was for language models, because every attempt spends real robot hours instead of data-centre compute. CFChelsea Finn12 Aug 2026LLMsreinforcement learning
• Without an execution environment, LLMs cannot properly act on their understanding of how to forward evolve a system. TLTobias Lütke21 Sept 2026LLMsreinforcement learning
• A verifiable task can be optimised by reinforcement learning until a neural network performs it extremely well. AKAndrej Karpathy17 Nov 2025reinforcement learning
• Capability does not generalise for free: a model that will move mountains on an agentic task still tells the same bad joke it told five years ago. AKAndrej Karpathy20 Mar 2026reinforcement learning

See this claim on the map