Kevin Patrick Murphy @sirbayes · 12 Aug 2026
**11/ Why the learned summary doesn't collapse** Self-supervised encoders (JEPA-style) risk **representational collapse** (s(y)→const), patched with stop-grad/EMA hacks. We avoid it *for free*: s_φ is trained by a **supervised** objective — predict (m,θ) — which anchors it.
4 303Views
28Likes
1Reposts
1Replies
0Quotes
8Bookmarks
Is that a lot?
0.67×vs this author's median6 395 views is typical
22Percentile for this authorof 18 recent posts
3.11×vs 10K–100K median1 384 views is typical
5.75%Reachviews ÷ followers
0.70%Engagement rateof viewers reacted