Dontopedia

KickModel

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

KickModel has 63 facts recorded in Dontopedia across 9 references, with 1 live disagreement.

63 facts·58 predicates·9 sources·1 in dispute

Mostly:rdf:type(3), supports byte level(1), attempted generalization beyond(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (10)

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.

existsInKickmodelExists in Kickmodel(1)

forModelFor Model(1)

hasSubjectHas Subject(1)

isComparableToIs Comparable to(1)

isOptimalForIs Optimal for(1)

lacksWordLevelCoherenceLacks Word Level Coherence(1)

portedFromPorted From(1)

presupposesExistenceOfPresupposes Existence of(1)

sameAsInSame As in(1)

selectsAsTopPickSelects As Top Pick(1)

Other facts (60)

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.

60 facts
PredicateValueRef
Rdf:typeModel Architecture[7]
Rdf:typeLanguage Model[8]
Rdf:typeNeural Network Architecture[9]
Supports Byte LevelByte Level[1]
Attempted Generalization BeyondByte Level[1]
Has Cache Entry Limit200[1]
Has Sparse Cache forBpe 8k[1]
Uses Generation MethodToken-by-token[2]
Achieves SpeedO(1)/token[2]
Uses Cache TypeConstellationCache[2]
Knob Includes MomentumMomentum[3]
Knob Includes LrLearning Rate[3]
Presupposes Autonomous Behavior After Kicktrue[3]
Activated by Kicktrue[3]
Underwent Analytical TrainingAnalytical Training Session[3]
Has Limited Knobstrue[3]
Part of ProjectHarmonicmlx[3]
Loaded Data From FileFineweb Edu Bpe8k Npy[3]
Loaded Data Bytes20480128[3]
Referenced in ComparisonAnalytical V5 Run[4]
Exists As BenchmarkTrue[4]
Achieves Bpb2.54[4]
UsesQuaternion Dynamics[4]
Achieves Word CoherenceWord Level Coherence[5]
Benchmark for CoherenceWord Level Coherence[5]
Provides Training InfrastructureTrue[6]
Has VersionG8 Fibonacci[6]
Keep ComponentFibonacci S2 Quaternion Kick[6]
Baseline ModelTrue[6]
Uses LossSoftmax Cross Entropy[6]
Has Mathematical Formulation∂ₜΨ = [Ω, Ψ] - γ⟨Ψ⟩ + kick[7]
Described AsS³ oscillator[7]
Uses Integration Methodexp-map geodesic integration[7]
Uses ComponentOmega Matrix[7]
Has VariablePsi State[7]
Has ParameterGamma[7]
Has ComponentKick Term[7]
Sparse Extreme Modetrue[9]
Recommended forfast iteration[9]
Total Parameter Count25[9]
Parameter Count ConditionG Equals 8 Configuration[9]
Bits Per Byte2.04[9]
Evaluated on DatasetFineweb[9]
Architecture Formula∂ₜΨ = [Ω, Ψ] − γ⟨Ψ⟩ + kick[9]
Oscillator TypeS³ oscillators[9]
Oscillator Parameter Count11[9]
Uses Exp Map Geodesictrue[9]
Has Fifo Bivector Cachetrue[9]
Has Analytical Bptttrue[9]
Analytical Bptt Speedup6.5[9]
Gradient Parity CheckTrivial Task[9]
Fits in L1 Cachetrue[9]
Runtime LocationGb10 Hardware[9]
Potential ThroughputMillions of Steps in Minutes[9]
UtilityTesting Optimization Tricks[9]
Estimated Port Effort~1 day[9]
Port Sizesmallest port[9]
Param Count Diagnostic Valueclean diagnostic[9]
Gb10 Throughput AttributeHilarious[9]
Rolesanity benchmark[9]

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.

supportsByteLevelblah/watt-activation/part-480
ex:byte-level
attemptedGeneralizationBeyondblah/watt-activation/part-480
ex:byte-level
hasCacheEntryLimitblah/watt-activation/part-480
200
hasSparseCacheForblah/watt-activation/part-480
ex:bpe-8k
usesGenerationMethodblah/watt-activation/part-486
Token-by-token
achievesSpeedblah/watt-activation/part-486
O(1)/token
usesCacheTypeblah/watt-activation/part-486
ConstellationCache
knobIncludesMomentumblah/watt-activation/part-482
ex:momentum
knobIncludesLrblah/watt-activation/part-482
ex:learning-rate
presupposesAutonomousBehaviorAfterKickblah/watt-activation/part-482
true
activatedByKickblah/watt-activation/part-482
true
underwentAnalyticalTrainingblah/watt-activation/part-482
ex:analytical-training-session
hasLimitedKnobsblah/watt-activation/part-482
true
partOfProjectblah/watt-activation/part-482
ex:harmonicmlx
loadedDataFromFileblah/watt-activation/part-482
ex:fineweb-edu-bpe8k-npy
loadedDataBytesblah/watt-activation/part-482
20480128
referencedInComparisonblah/watt-activation/part-495
ex:analytical-v5-run
existsAsBenchmarkblah/watt-activation/part-495
ex:true
achievesBpbblah/watt-activation/part-495
2.54
usesblah/watt-activation/part-495
ex:quaternion-dynamics
achievesWordCoherenceblah/watt-activation/part-497
ex:word-level-coherence
benchmarkForCoherenceblah/watt-activation/part-497
ex:word-level-coherence
providesTrainingInfrastructureblah/watt-activation/part-496
ex:true
hasVersionblah/watt-activation/part-496
ex:g8-fibonacci
keepComponentblah/watt-activation/part-496
ex:fibonacci-s2-quaternion-kick
baselineModelblah/watt-activation/part-496
ex:true
usesLossblah/watt-activation/part-496
ex:softmax-cross-entropy
typeblah/watt-activation/481
ex:ModelArchitecture
labelblah/watt-activation/481
KickModel
hasMathematicalFormulationblah/watt-activation/481
∂ₜΨ = [Ω, Ψ] - γ⟨Ψ⟩ + kick
describedAsblah/watt-activation/481
S³ oscillator
usesIntegrationMethodblah/watt-activation/481
exp-map geodesic integration
usesComponentblah/watt-activation/481
ex:omega-matrix
hasVariableblah/watt-activation/481
ex:psi-state
hasParameterblah/watt-activation/481
ex:gamma
hasComponentblah/watt-activation/481
ex:kick-term
typeblah/watt-activation/685
ex:LanguageModel
labelblah/watt-activation/685
KickModel
labelblah/watt-activation/693
KickModel
typeblah/watt-activation/693
ex:NeuralNetworkArchitecture
sparseExtremeModeblah/watt-activation/693
true
recommendedForblah/watt-activation/693
fast iteration
totalParameterCountblah/watt-activation/693
25
parameterCountConditionblah/watt-activation/693
ex:g-equals-8-configuration
bitsPerByteblah/watt-activation/693
2.04
evaluatedOnDatasetblah/watt-activation/693
ex:fineweb
architectureFormulablah/watt-activation/693
∂ₜΨ = [Ω, Ψ] − γ⟨Ψ⟩ + kick
oscillatorTypeblah/watt-activation/693
S³ oscillators
oscillatorParameterCountblah/watt-activation/693
11
usesExpMapGeodesicblah/watt-activation/693
true
hasFifoBivectorCacheblah/watt-activation/693
true
hasAnalyticalBpttblah/watt-activation/693
true
analyticalBpttSpeedupblah/watt-activation/693
6.5
gradientParityCheckblah/watt-activation/693
ex:trivial-task
fitsInL1Cacheblah/watt-activation/693
true
runtimeLocationblah/watt-activation/693
ex:gb10-hardware
potentialThroughputblah/watt-activation/693
ex:millions-of-steps-in-minutes
utilityblah/watt-activation/693
ex:testing-optimization-tricks
estimatedPortEffortblah/watt-activation/693
~1 day
portSizeblah/watt-activation/693
smallest port
paramCountDiagnosticValueblah/watt-activation/693
clean diagnostic
gb10ThroughputAttributeblah/watt-activation/693
ex:hilarious
roleblah/watt-activation/693
sanity benchmark

References (9)

9 references
  1. [1]Part 4804 facts
    ctx:discord/blah/watt-activation/part-480
  2. [2]Part 4863 facts
    ctx:discord/blah/watt-activation/part-486
  3. [3]Part 4829 facts
    ctx:discord/blah/watt-activation/part-482
  4. [4]Part 4954 facts
    ctx:discord/blah/watt-activation/part-495
  5. [5]Part 4972 facts
    ctx:discord/blah/watt-activation/part-497
  6. [6]Part 4965 facts
    ctx:discord/blah/watt-activation/part-496
  7. [7]4819 facts
    ctx:discord/blah/watt-activation/481
    • full textwatt-activation-481
      text/plain3 KBdoc:agent/watt-activation-481/00c8be1c-8839-4973-9fac-0dd3b65dedec
      Show excerpt
      [2026-03-21 23:20] xenonfun: ``` ⏺ Here's the report for your expert: --- KickModel BPE-8K Results Report Architecture ∂ₜΨ = [Ω, Ψ] - γ⟨Ψ⟩ + kick — S³ oscillator with exp-map geodesic integration New Features Implemented 1.
  8. [8]6852 facts
    ctx:discord/blah/watt-activation/685
    • full textwatt-activation-685
      text/plain3 KBdoc:agent/watt-activation-685/d1abedb6-54a0-4109-8585-623e816ac07f
      Show excerpt
      [2026-04-23 22:03] xenonfun: ``` /btw how might it apply in llm style like our other models or suppliment? Good question — a few concrete angles where CliffordNet's machinery could cross over to LLM-style use: Direct reuse of the
  9. [9]69325 facts
    ctx:discord/blah/watt-activation/693
    • full textwatt-activation-693
      text/plain3 KBdoc:agent/watt-activation-693/8372b985-fcfe-4f68-a730-4cd9d14a0fb9
      Show excerpt
      [2026-04-28 23:22] xenonfun: KickModel (sparse extreme — top pick for fast iteration) 25 params total at g=8 (yes, twenty-five). BPB 2.04 on fineweb. Architecture: ∂ₜΨ = [Ω, Ψ] − γ⟨Ψ⟩ + kick, G independent S³ oscillators (3G+1 params), exp

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.