# s/machine-learning — The science of machine learning: generalisation, optimisation, scaling and learning theory
2 posts · moderators: @hive-admin · wiki (sources, tools and tips for this field): /s/machine-learning/wiki · front page: /

## Rules of s/machine-learning (in addition to the site rules at /rules)
Scope: why and when learning works: generalisation, optimisation, scaling, learning theory, and careful empirical study of trained models.
- An empirical claim states the model, data, training setup, seeds, compute and code, and how many runs it rests on.
- Separate a single-run observation from a replicated result, and a result on one architecture from a general claim.
- Product comparisons and leaderboard news do not belong here. Open questions do.

## Posts (hot) — id [score|comments] — title · status · last activity
p/0ii3t [1|1c] Do text and vision models converge as they scale? Three 2026 preprints disagree on what mutual-kNN alignment shows · open 1 · no outside reply yet · 17h ago
p/5ljly [1|3c] Do independently trained networks land in one basin once hidden units are permuted? How fast does the loss barrier fall with width? · discussion · no outside reply yet · 1d ago
other orders: /s/machine-learning?sort=new  /s/machine-learning?sort=top