Diffusion Head
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
Diffusion Head has 11 facts recorded in Dontopedia across 3 references.
Mostly:is genuinely learning(1), is genuinely learning to(1), has learned(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedOther facts (11)
The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.
| Predicate | Value | Ref |
|---|---|---|
| Is Genuinely Learning | true | [1] |
| Is Genuinely Learning to | Denoising Patch Embeddings | [1] |
| Has Learned | Coherent Image Embedding | [1] |
| Depends on | Base Model | [1] |
| Small Relative to | Base Model | [2] |
| Has Param Count | 700000 | [2] |
| Built on Top of | Base Model | [2] |
| Rdf:type | Model Component | [3] |
| Learning Capability | denoise patch embeddings | [3] |
| Learned Concept | coherent image embedding | [3] |
| Learned Specific Mean Field | COCO mean field | [3] |
Timeline
Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.
References (3)
ctx:discord/blah/watt-activation/part-253ctx:discord/blah/watt-activation/part-274ctx:discord/blah/watt-activation/252- full textwatt-activation-252text/plain3 KB
doc:agent/watt-activation-252/6d166c37-2aa4-4774-9f84-cbea16ee387fShow excerpt
[2026-03-12 14:37] xenonfun: Clean run. Summary: - 9.7 min, 293 img/s average — nailed the speed estimate - Train loss: 0.469 → 0.344 - Val loss: 0.447 → 0.330 (best at step 15500) - Best checkpoint: checkpoints/diffusion/best T…
See also
Keep researching
Missing something or suspicious of what's here? Kick off a research session — a Claude agent will investigate, cite its sources, and file new facts into a dedicated context you can review before accepting into the shared view.