Dontopedia
Explore

Softmax in Diversity Loss

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)

Softmax in Diversity Loss has 2 facts recorded in Dontopedia across 2 references.

2 facts·2 predicates·2 sources
Maturity scale raw canonical shape-checked rule-derived certified

Is Not PrincipledisNotPrincipled

Rdf:typerdf:type

Inbound mentions (3)

Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. "X motherOf this subject" — and answer questions the forward facts can't. Grouped by predicate.

addressesAddresses(1)

asksWhyAsks Why(1)

questionsPrincipledNatureOfQuestions Principled Nature of(1)

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.

2 undated
—
isNotPrincipledblah/watt-activation/part-282
ex:for-s-d-1
—
typeblah/watt-activation/280
ex:DesignDecision

References (2)

2 references
  1. [1]Part 2821 fact
    customctx:discord/blah/watt-activation/part-282
  2. [2]watt-activation-2801 fact
    customctx:discord/blah/watt-activation/280
    • full textwatt-activation-280
      text/plain2 KBdoc:agent/watt-activation-280/4706b1d1-971f-4a85-9e69-f1be1b2b13c8
      Show excerpt
      [2026-03-13 23:07] xenonfun: ``` The fundamental finding: The codebook doesn't carry meaningful information. Loss is identical with or without VQ. Utilization is 1-2% everywhere. The model learns byte prediction entirely through the b…

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.

↑