"By comparing LRMs with their standard LLM counterparts under equivalent inference compute, we identify three performance regimes: (1) low- complexity tasks where standard models surprisingly outperform LRMs, (2) medium-complexity tasks where additional thinking in LRMs demonstrates advantage, and (3) high-complexity tasks where both models experience complete collapse."
https://machinelearning.apple.com/research/illusion-of-thinking
#AI #LLM #Reasoning
models outperform both their high-complexity compute, under medium-complexity (3) three "By complexity low- equivalent #Reasoning comparing (1) where thinking standard complete experience models demonstrates #LLM tasks where inference performance tasks counterparts in with LRMs additional collapse."
https://machinelearning.apple.com/research/illusion-of-thinking
#AI where identify (2) we LRMs, regimes: and LRMs standard tasks surprisingly advantage, LLM
