LLMs still fail at temporal reasoning, and a hierarchical HMM is proposed for extreme long contexts

beffjezos · x · 2026-07-27

A post argues that whoever figures out a temporal hierarchical hidden Markov model for LLMs over extreme long contexts will “print,” in response to a complaint that frontier models still fail at temporal reasoning: they produce giant essays, misunderstand time, and show a strong recency bias.

Original post →

More from AGI Musings

AGI Musings channel →