Dontopedia

Key Finding 4

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

Key Finding 4 has 21 facts recorded in Dontopedia across 7 references, with 2 live disagreements.

21 facts·13 predicates·7 sources·2 in dispute

Mostly:rdf:type(4), claims stability(1), described in(1)

Maturity scale raw canonical shape-checked rule-derived certified

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.

hasFindingHas Finding(1)

hasKeyFindingHas Key Finding(1)

mentionsFindingMentions Finding(1)

Other facts (16)

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.

16 facts
PredicateValueRef
Rdf:typeClaim[2]
Rdf:typeKey Finding[3]
Rdf:typeKey Finding[4]
Rdf:typeClaim[5]
Claims StabilityTraining[1]
Described inLog Entry 2026 03 17 15 27[3]
Persistence ConditionCurriculum Removal[4]
Claimed byXenonfun[5]
References Correlation Range0.43-0.55[5]
Implies CauseFewer Groups[5]
States Stabilitystable[6]
Reports Throughput96K tok/s[6]
Uses TechnologyRayon Parallelism[6]
Runs on HardwareM3 Ultra[6]
Assesses ArchitectureHybrid Fixed Beta[7]
Identifies WeaknessBase Model[7]

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.

21 undated
—
claimsStabilityblah/watt-activation/part-605
ex:training
—
typeblah/watt-activation/292
ex:Claim
—
labelblah/watt-activation/355
Key Finding 4
—
typeblah/watt-activation/355
ex:KeyFinding
—
describedInblah/watt-activation/355
ex:log-entry-2026-03-17-15-27
—
typeblah/watt-activation/360
ex:KeyFinding
—
labelblah/watt-activation/360
Finding 4: Retrieval persistence
—
persistenceConditionblah/watt-activation/360
ex:curriculum-removal
—
typeblah/watt-activation/393
ex:Claim
—
labelblah/watt-activation/393
QPSK models have lower r
—
claimedByblah/watt-activation/393
ex:xenonfun
—
referencesCorrelationRangeblah/watt-activation/393
0.43-0.55
—
impliesCauseblah/watt-activation/393
ex:fewer-groups
—
labelblah/watt-activation/602
Key Finding 4
—
statesStabilityblah/watt-activation/602
stable
—
reportsThroughputblah/watt-activation/602
96K tok/s
—
usesTechnologyblah/watt-activation/602
ex:rayon-parallelism
—
runsOnHardwareblah/watt-activation/602
ex:m3-ultra
—
labelblah/watt-activation/608
Correct architecture, weak base model
—
assessesArchitectureblah/watt-activation/608
ex:hybrid-fixed-beta
—
identifiesWeaknessblah/watt-activation/608
ex:base-model

References (7)

7 references
  1. [1]Part 6051 fact
    ctx:discord/blah/watt-activation/part-605
  2. [2]2921 fact
    ctx:discord/blah/watt-activation/292
    • full textwatt-activation-292
      text/plain2 KBdoc:agent/watt-activation-292/126da404-b457-4703-9947-909873e9c37a
      Show excerpt
      [2026-03-14 03:52] xenonfun: ``` ⏺ There it is: LOHE OPTIMIZER STEP BREAKDOWN (492.5 ms total, 66.5K tok/s) ├── Forward + capture: 143.4 ms (29%) — same as RotAdamW forward ├── Block updates: 344.3 ms (70%) ← THE …
  3. [3]3553 facts
    ctx:discord/blah/watt-activation/355
    • full textwatt-activation-355
      text/plain3 KBdoc:agent/watt-activation-355/e62c81a8-1082-4c07-b675-3759a8600d0e
      Show excerpt
      [2026-03-17 15:27] xenonfun: ``` Key findings: 1. Depth scaling is smooth and strong: BPB drops monotonically 3.00→2.53 from D=6→D=32. DC@16 rises 72%→91%. 2. Retrieval reach = 128 for ALL configs — every model retrieves across the f…
  4. [4]3603 facts
    ctx:discord/blah/watt-activation/360
    • full textwatt-activation-360
      text/plain2 KBdoc:agent/watt-activation-360/2d704135-eed2-4a3e-9603-3e55129dda47
      Show excerpt
      [2026-03-17 19:08] xenonfun: ``` --- Session Summary Architecture validated - Mercury-delay-line field transport with oscillator transduction - Depth is the primary scaling axis (not K) - Retrieval is distance-invariant (DC@16 …
  5. [5]3935 facts
    ctx:discord/blah/watt-activation/393
    • full textwatt-activation-393
      text/plain3 KBdoc:agent/watt-activation-393/d9206d9a-8fe3-4b8a-9bba-45db3179f660
      Show excerpt
      [2026-03-19 03:51] xenonfun: ⏺ Excellent results! Almost all done. Here's the encoding comparison with proper LR: ``` ┌─────────┬───────────┬───────────┬──────────┬───────┬───────┬────────┐ │ Config │ Encoding │ Block │ BPB @500 │…
  6. [6]6025 facts
    ctx:discord/blah/watt-activation/602
    • full textwatt-activation-602
      text/plain3 KBdoc:agent/watt-activation-602/b35f7366-81fb-47bd-8e2b-a00aa620e54a
      Show excerpt
      [2026-04-10 04:02] xenonfun: ``` First Results Summary ┌────────────────────┬─────────┬───────┬───────────────────────────┐ │ Variant │ BPB @5K │ tok/s │ Delta Memory Active? │ ├────────────────────┼─────────┼──────…
  7. [7]6083 facts
    ctx:discord/blah/watt-activation/608
    • full textwatt-activation-608
      text/plain2 KBdoc:agent/watt-activation-608/a9cc9bc2-b034-450f-bf85-dcb33eeaecc4
      Show excerpt
      [2026-04-10 19:18] xenonfun: at 85% test coverage on library, few more improvement and enable the swarm downloading, server becomes just a seeding node mostly. ✶ Improving handler test coverage… ⎿  ◼ Improve handler.rs coverage (49% → 70%…

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.

↑