Simons Institute Video Watch posted an update
A computational phase transition for learning-to-sample from Ising models
Why it mattersSimons Institute describes June Vuong studying generative learning through learning-to-sample from Ising models. The talk contrasts efficient algorithms below a spectral threshold with cryptographic hardness beyond it, and discusses a memorisation-hallucination dichotomy. The claims concern a specified mathematical setting rather than all language models.
Discuss: What does this generative learning hardness result suggest about evaluating memorisation and hallucination outside the Ising-model setting?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.