Random Init

as donto believed on

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

Random Init has 10 facts recorded in Dontopedia across 7 references, with 1 live disagreement.

10 facts·9 predicates·7 sources·1 in dispute

Mostly:sanity check value for(2), precedes(1), implies decoherence(1)

Inbound mentions (8)

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.

`associatedWith`Associated With(1)

Disordered Phaseex:disordered-phase

`assumesRandomInitialization`Assumes Random Initialization(1)

Test 1ex:test-1

`at`At(1)

Rex:r

`condition`Condition(1)

Gate 0.54 at Random Initex:gate-0.54-at-random-init

`expectedAt`Expected at(1)

Cross Layer Decoherenceex:cross-layer-decoherence

`noisyAtInit`Noisy at Init(1)

Featuresex:features

`notFrom`Not From(1)

Ω Iex:ω_i

`startsWith`Starts With(1)

From Scratch Trainingex:from-scratch-training

Other facts (10)

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.

10 facts

Predicate	Value	Ref
Sanity Check Value for	r	[6]
Sanity Check Value for	beta	[6]
Precedes	2k Iters	[1]
Implies Decoherence	Cross Layer	[2]
Has Gate Value	0.54	[3]
Has R Value	≈0.16	[3]
Is Slower Than	Fine Tune Start	[3]
Produces Disordered Regime	null	[4]
Causes Noise Issue for	Lohe Optimizer	[5]
Rdf:type	Training Phase	[7]

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.

10 undated

—

precedesblah/watt-activation/part-75

ex:2k-iters

—

impliesDecoherenceblah/watt-activation/part-187

ex:cross-layer

—

hasGateValueblah/watt-activation/part-199

0.54

—

hasRValueblah/watt-activation/part-199

≈0.16

—

isSlowerThanblah/watt-activation/part-199

ex:fine-tune-start

—

producesDisorderedRegimeblah/watt-activation/part-185

null

—

causesNoiseIssueForblah/watt-activation/part-205

ex:lohe-optimizer

—

sanityCheckValueForblah/watt-activation/185

—

sanityCheckValueForblah/watt-activation/185

beta

—

typeblah/watt-activation/198

ex:TrainingPhase

References (7)

7 references

[1]Part 751 fact
ctx:discord/blah/watt-activation/part-75
[2]Part 1871 fact
ctx:discord/blah/watt-activation/part-187
[3]Part 1993 facts
ctx:discord/blah/watt-activation/part-199
[4]Part 1851 fact
ctx:discord/blah/watt-activation/part-185
[5]Part 2051 fact
ctx:discord/blah/watt-activation/part-205
[6]1852 facts
ctx:discord/blah/watt-activation/185
- full textwatt-activation-185
  text/plain3 KBdoc:agent/watt-activation-185/fcee3d0e-68fc-4f0b-ad9d-157115bfbade
  Show excerpt
  [2026-03-10 02:38] lisamegawatts: i think this part is wrong, it should not init at zero they need frequencies. i also don't think it needs gelu, the linear error is inherently in the sphere; The Lohe sphere model, a generalization of the K…
[7]1981 fact
ctx:discord/blah/watt-activation/198
- full textwatt-activation-198
  text/plain2 KBdoc:agent/watt-activation-198/ea2dbbd7-7d1c-4ba2-9281-56322a048288
  Show excerpt
  [2026-03-10 06:11] xenonfun: ``` ⏺ --- Summary #1 — features tensor restored One line in lohe_ffn.py. The optimizer already had the branch ready (if 'features' in metrics:). What activates now: - W_out[:, 32..92] (weighted harmon…

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.

Random Init

Inbound mentions (8)

associatedWithAssociated With(1)

assumesRandomInitializationAssumes Random Initialization(1)

atAt(1)

conditionCondition(1)

expectedAtExpected at(1)

noisyAtInitNoisy at Init(1)

notFromNot From(1)

startsWithStarts With(1)