Open-world AI: a Tale of Three Requirements

A new arXiv paper (2610.20782v1) argues that open-world AI requires three properties — inference-time learning, a rich representation, and predictive disentanglement — and reports that language models up to 8B parameters meeting these requirements improve as data from related tasks accumulates while the rest degrade. The analysis, which models tasks as sharing latent factors and decomposes the training objective, further predicts the need for a consistent inner learner and the benefit of long contexts spanning related tasks. Neither scale nor layer-type interleaving repaired the degradation in models that failed the requirements.

arXiv:2610.20782v1 Announce Type: new Abstract: The closed world asks AI for specialized competence and lets it manufacture the iid data its theory assumes: a fixed criterion licenses collecting examples until the task is covered. The open world asks for versatility --- taking up tasks a machine was not built for, from few examples and with little prior from a human designer --- and offers data that is grouped, not shuffled. Its tasks are not known in advance, so a machine must do inference-time learning 1st requirement . Modeling tasks as sharing latent factors and decomposing the training objective, we derive two more: a rich representation , retaining more than iid generalization requires, and a predictive disentanglement that aligns inputs and outputs into matched low-dimensional groups. The analysis further predicts the need for a consistent inner learner, and the benefit of long contexts spanning related tasks. We test these claims on language models up to 8B, asking whether a model has a property qualitative rather than how much it wins quantitative . Models that meet the requirements improve as data from related tasks accumulates; the rest degrade, and neither scale nor layer-type interleaving repairs this.