#cognitivelimits — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #cognitivelimits, aggregated by home.social.
-
Do reasoning models actually “think”?
A new Apple research paper shows: Large Reasoning Models (LRMs) like Claude or DeepSeek can reason — up to a point. But as complexity rises, accuracy collapses. Surprisingly, models think less when problems get harder.
📄 https://machinelearning.apple.com/research/illusion-of-thinking
#AI #Reasoning #LLM #GenerativeAI #CognitiveLimits #Claude3 #DeepSeek -
The Illusion of Thinking: Understanding the Limitations of Reasoning LLMs [pdf]
https://ml-site.cdn-apple.com/papers/the-illusion-of-thinking.pdf
#HackerNews #IllusionOfThinking #ReasoningLLMs #AIResearch #CognitiveLimits #PDFInsights