#Llm-Reasoning
2 notas · última:
The Illusion of the Illusion of Thinking
Anthropic and Open Philanthropy researchers argue that a widely cited 'reasoning collapse' study measured token limits and broken benchmarks, not a failure of model reasoning.
The Illusion of Thinking
Apple argues that frontier large reasoning models improve on some reasoning benchmarks, but their accuracy collapses past certain problem complexities, their reasoning effort can fall as complexity rises, and they show limits in exact computation.