ResonantWireLM

as donto believed on

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

ResonantWireLM has 212 facts recorded in Dontopedia across 21 references, with 16 live disagreements.

212 facts·172 predicates·21 sources·16 in dispute

Mostly:has hyperparameter(9), has readout dim(5), rdf:type(4)

Inbound mentions (32)

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.

`partOfModel`Part of Model(12)

Block 0 Diagnosticsex:block-0-diagnostics
Block 10 Diagnosticsex:block-10-diagnostics
Block 11 Diagnosticsex:block-11-diagnostics
Block 1 Diagnosticsex:block-1-diagnostics
Block 2 Diagnosticsex:block-2-diagnostics
Block 3 Diagnosticsex:block-3-diagnostics
Block 4 Diagnosticsex:block-4-diagnostics
Block 5 Diagnosticsex:block-5-diagnostics
Block 6 Diagnosticsex:block-6-diagnostics
Block 7 Diagnosticsex:block-7-diagnostics
Block 8 Diagnosticsex:block-8-diagnostics
Block 9 Diagnosticsex:block-9-diagnostics

`partOf`Part of(2)

Groupsex:groups
Rotorsex:rotors

`advocatesScaling`Advocates Scaling(1)

Xenonfunex:xenonfun

`associatedWithResonantWireLm`Associated With Resonant Wire Lm(1)

Rotor Modelex:rotor-model

`describedPerBlockPerToken`Described Per Block Per Token(1)

Scalingex:scaling

`doesNotApplyTo`Does Not Apply to(1)

Sphere Projection Diagnosticsex:sphere-projection-diagnostics

`existsInContextOf`Exists in Context of(1)

Rotorex:rotor

`existsInResonantwirelm`Exists in Resonantwirelm(1)

Physicsex:physics

`forModel`For Model(1)

Entity Curriculum Resonantwirelmex:entity-curriculum-resonantwirelm

`hasModel`Has Model(1)

Generation Runex:generation-run

`includesResonantWireLM`Includes Resonant Wire Lm(1)

Today Arcex:today-arc

`inquiresAboutMemoryRequirements`Inquires About Memory Requirements(1)

Xenonfunex:xenonfun

`isBottleneck`Is Bottleneck(1)

Rotor Parallel Scanex:rotor-parallel-scan

`isCheckpointFor`Is Checkpoint for(1)

Gated Rotor 2k Checkpointex:gated-rotor-2k-checkpoint

`isStandalone`Is Standalone(1)

Resonantwirelm Evaluationex:resonantwirelm-evaluation

`isUnmeasured`Is Unmeasured(1)

Physicsex:physics

`offersToTrain`Offers to Train(1)

Xenonfunex:xenonfun

`postedModelInfo`Posted Model Info(1)

Xenonfunex:xenonfun

`specificTo`Specific to(1)

Rotorsex:rotors

`statesThat`States That(1)

Message 2ex:message-2

Other facts (206)

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.

206 facts

Predicate	Value	Ref
Has Hyperparameter	G=8	[15]
Has Hyperparameter	H=4	[15]
Has Hyperparameter	L=6	[15]
Has Hyperparameter	heads=4	[15]
Has Hyperparameter	k_probes=4	[15]
Has Hyperparameter	constellation_temp=0.254	[17]
Has Hyperparameter	readout_dim=93	[17]
Has Hyperparameter	constellation_temp=0.100	[21]
Has Hyperparameter	readout_dim=93	[21]
Has Readout Dim	93	[5]
Has Readout Dim	93	[9]
Has Readout Dim	93	[10]
Has Readout Dim	64	[13]
Has Readout Dim	93	[18]
Rdf:type	Model Architecture	[15]
Rdf:type	Model Architecture	[19]
Rdf:type	Model Architecture	[20]
Rdf:type	Model	[21]
Has Params	33881	[2]
Has Params	27129	[10]
Has Params	49506	[13]
Has Constellation Temp	0.109	[5]
Has Constellation Temp	0.101	[9]
Has Constellation Temp	0.1	[10]
Has Fundamentally Different Block Structure Than	Harmonicgpt	[1]
Has Fundamentally Different Block Structure Than	Wirelm	[1]
Has Highest Throughput	All Tested Models	[2]
Has Highest Throughput	true	[15]
Loaded in Ms	54	[5]
Loaded in Ms	64	[9]
Has G	8	[5]
Has G	8	[9]
Has Param Count	27057	[5]
Has Param Count	30297	[9]
Has H	2	[5]
Has H	2	[9]
Has L	12	[5]
Has L	24	[9]
Has Heads	4	[5]
Has Heads	4	[9]
Has Num Heads	2	[6]
Has Num Heads	4	[13]
Uses	Helmholtz Fiber	[8]
Uses	Givens Coupling	[8]
Has Characteristic	Pure Rotation	[16]
Has Characteristic	Geometry	[16]
Lacks Concept	Snr	[16]
Lacks Concept	Bandwidth	[16]
Load Time	56	[17]
Load Time	52	[21]
Best Validation Step	500	[17]
Best Validation Step	9500	[21]
Has Memory Size	0.11	[18]
Has Memory Size	0.11	[21]
Contrasts With	Wirelm Block Structure	[1]
Should Be Wired Into	Harmonicmlx Trainer Py	[1]
Commits to Being	Valid Model Type	[1]
Lacks Blocks With	Ffn Last Metrics	[1]
Has Own Block Type	True	[1]
Depends on	Own Block Type	[1]
Exists As Ready to Train	null	[2]
Requires Simple Training Loop	Standard Trainer	[2]
Is Small	null	[2]
Architecture Works	true	[2]
Commits to Learning Structure	null	[2]
Learns Real Sequential Structure	Dataset	[2]
Loaded From Checkpoint	checkpoints/resonant_wire/best	[2]
Lacks Dmodel	D Model	[3]
Lacks Residual Stream	Residual Stream	[3]
Lacks Snr Concept	Snr	[3]
Has Mode Power	Mode Power	[3]
Exists	Entity	[3]
Commits to Physics Analogs	Physical Interpretation	[3]
Computes Dft Mode Power Distribution	Dft Mode Power	[3]
Computes Graph Harmonics	Graph Harmonics	[3]
Computes R	R	[3]
Is Pure Rotation and Geometry	Current Implementation	[3]
Presupposes Rotation	Rotations	[3]
Has R	R	[3]
Is Not Measuring Physics	Physics	[3]
Described As Pure Rotation Geometry	Current State	[3]
Differs From Main Trainer	Main Trainer	[3]
Presupposes Geometry	Geometry	[3]
Has Rotor Quaternions	Rotor Quaternions	[3]
Lacks Attention	Attention Mechanism	[3]
Lacks Bandwidth Concept	Bandwidth	[3]
Known to Compute R	We Compute	[3]
Uses Adam Optimizer	Adam	[3]
Scales From Params	8500000	[4]
Achieves Bpb	4.06	[4]
Achieves Tok Per Sec	41000	[4]
Scales to Params	27000	[4]
Optimizes Bandwidth Usage	Full Bandwidth	[4]
Is Healthy	Training Dynamics	[4]
References Mamba	true	[5]
Achieved Best Val Bpb at Step	3000	[5]
Has Model Name	ResonantWireLM	[5]
Generated Gibberish	Model Output Gibberish	[5]
Achieved Best Val Bpb	3.675	[5]
Has Memory Mb	0.11	[5]

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.

212 undated

—

contrastsWithblah/watt-activation/part-387

ex:wirelm-block-structure

—

hasFundamentallyDifferentBlockStructureThanblah/watt-activation/part-387

ex:harmonicgpt

—

shouldBeWiredIntoblah/watt-activation/part-387

ex:harmonicmlx-trainer-py

—

commitsToBeingblah/watt-activation/part-387

ex:valid-model-type

—

lacksBlocksWithblah/watt-activation/part-387

ex:ffn-last-metrics

—

hasOwnBlockTypeblah/watt-activation/part-387

ex:true

—

hasFundamentallyDifferentBlockStructureThanblah/watt-activation/part-387

ex:wirelm

—

dependsOnblah/watt-activation/part-387

ex:own-block-type

—

existsAsReadyToTrainblah/watt-activation/part-386

null

—

requiresSimpleTrainingLoopblah/watt-activation/part-386

ex:standard-trainer

—

hasParamsblah/watt-activation/part-386

33881

—

isSmallblah/watt-activation/part-386

null

—

architectureWorksblah/watt-activation/part-386

true

—

commitsToLearningStructureblah/watt-activation/part-386

null

—

learnsRealSequentialStructureblah/watt-activation/part-386

ex:dataset

—

loadedFromCheckpointblah/watt-activation/part-386

checkpoints/resonant_wire/best

—

hasHighestThroughputblah/watt-activation/part-386

ex:all-tested-models

—

lacksDmodelblah/watt-activation/part-394

ex:d-model

—

lacksResidualStreamblah/watt-activation/part-394

ex:residual-stream

—

lacksSnrConceptblah/watt-activation/part-394

ex:snr

—

hasModePowerblah/watt-activation/part-394

ex:mode-power

—

existsblah/watt-activation/part-394

ex:entity

—

commitsToPhysicsAnalogsblah/watt-activation/part-394

ex:physical-interpretation

—

computesDftModePowerDistributionblah/watt-activation/part-394

ex:dft-mode-power

—

computesGraphHarmonicsblah/watt-activation/part-394

ex:graph-harmonics

—

computesRblah/watt-activation/part-394

ex:r

—

isPureRotationAndGeometryblah/watt-activation/part-394

ex:current-implementation

—

presupposesRotationblah/watt-activation/part-394

ex:rotations

—

hasRblah/watt-activation/part-394

ex:r

—

isNotMeasuringPhysicsblah/watt-activation/part-394

ex:physics

—

describedAsPureRotationGeometryblah/watt-activation/part-394

ex:current-state

—

differsFromMainTrainerblah/watt-activation/part-394

ex:main-trainer

—

presupposesGeometryblah/watt-activation/part-394

ex:geometry

—

hasRotorQuaternionsblah/watt-activation/part-394

ex:rotor-quaternions

—

lacksAttentionblah/watt-activation/part-394

ex:attention-mechanism

—

lacksBandwidthConceptblah/watt-activation/part-394

ex:bandwidth

—

knownToComputeRblah/watt-activation/part-394

ex:we-compute

—

usesAdamOptimizerblah/watt-activation/part-394

ex:adam

—

scalesFromParamsblah/watt-activation/part-396

8500000

—

achievesBPBblah/watt-activation/part-396

4.06

—

achievesTokPerSecblah/watt-activation/part-396

41000

—

scalesToParamsblah/watt-activation/part-396

27000

—

optimizesBandwidthUsageblah/watt-activation/part-396

ex:full-bandwidth

—

isHealthyblah/watt-activation/part-396

ex:training-dynamics

—

hasReadoutDimblah/watt-activation/part-400

—

referencesMambablah/watt-activation/part-400

true

—

achievedBestValBPBAtStepblah/watt-activation/part-400

3000

—

hasModelNameblah/watt-activation/part-400

ResonantWireLM

—

loadedInMsblah/watt-activation/part-400

—

generatedGibberishblah/watt-activation/part-400

ex:model-output-gibberish

—

achievedBestValBPBblah/watt-activation/part-400

3.675

—

hasGblah/watt-activation/part-400

—

hasMemoryMBblah/watt-activation/part-400

0.11

—

hasParamCountblah/watt-activation/part-400

27057

—

smallModelblah/watt-activation/part-400

27057

—

hasHblah/watt-activation/part-400

—

hasLblah/watt-activation/part-400

—

underwentGenerationblah/watt-activation/part-400

ex:generation-session-1

—

hasHeadsblah/watt-activation/part-400

—

hasConstellationTempblah/watt-activation/part-400

0.109

—

presupposesExistenceOfBlocksblah/watt-activation/part-401

null

—

modelSizeInMBblah/watt-activation/part-401

0.11

—

handlesLongerSequencesWithoutArchChangeblah/watt-activation/part-401

null

—

groupHeadProductEqualsblah/watt-activation/part-401

—

supportsTrainingSeqRangeblah/watt-activation/part-401

1K-4K

—

requiresNoArchitectureChangesForLongSeqblah/watt-activation/part-401

null

—

implicatesScalabilityTo16kblah/watt-activation/part-401

null

—

hasNumLayersblah/watt-activation/part-401

—

hasNumHeadsblah/watt-activation/part-401

—

hasNumGroupsblah/watt-activation/part-401

—

activationsMemoryAtSeq4kBatchSize64blah/watt-activation/part-401

768

—

existsAsLanguageModelblah/watt-activation/part-401

null

—

capableOfSeq16kWith48mbblah/watt-activation/part-401

null

—

presupposesTokensInSequenceblah/watt-activation/part-401

null

—

currentlyUsesSequenceSizeblah/watt-activation/part-401

256

—

lowModelSizeAdvantageblah/watt-activation/part-401

null

—

memoryAndComputeScaleWithblah/watt-activation/part-401

ex:sequence-length

—

benefitsFromDepthScalingblah/watt-activation/part-430

null

—

hasConfirmedDepthScalingblah/watt-activation/part-430

true

—

usesblah/watt-activation/part-424

ex:helmholtz-fiber

—

hasblah/watt-activation/part-424

ex:different-dynamics

—

usesblah/watt-activation/part-424

ex:givens-coupling

—

avoidsDynH4Flawsblah/watt-activation/part-424

true

—

hasNoblah/watt-activation/part-424

ex:lohe-sync-step

—

isCustomModelblah/watt-activation/part-432

true

—

achievesStrongValidationblah/watt-activation/part-432

ex:bpb-3-0568

—

hasReadoutDimblah/watt-activation/part-432

—

hasConstellationTempblah/watt-activation/part-432

0.101

—

hasHeadsblah/watt-activation/part-432

—

hasHblah/watt-activation/part-432

—

hasLblah/watt-activation/part-432

—

achievedBestValAtStepblah/watt-activation/part-432

10000

—

loadedInMsblah/watt-activation/part-432

—

hasBestValBPBblah/watt-activation/part-432

3.0568

—

hasParamCountblah/watt-activation/part-432

30297

—

hasSizeMBblah/watt-activation/part-432

0.12

—

hasGblah/watt-activation/part-432

—

bestValStepblah/watt-activation/part-431

9500

—

hasConfigHeadsblah/watt-activation/part-431

—

hasReadoutDimblah/watt-activation/part-431

—

loadTimeMsblah/watt-activation/part-431

—

referencesPriorWorkblah/watt-activation/part-431

ex:resonantwirelm-config

—

hasConfigGblah/watt-activation/part-431

—

hasConfigLblah/watt-activation/part-431

—

hasMemoryUsageblah/watt-activation/part-431

0.11MB

—

bestValBpbblah/watt-activation/part-431

3.1401

—

hasParamsblah/watt-activation/part-431

27129

—

hasConfigHblah/watt-activation/part-431

—

hasConstellationTempblah/watt-activation/part-431

0.1

—

usesCacheTypeblah/watt-activation/part-486

RotorCache + ReadoutCache

—

differsFromOthersInSpeedblah/watt-activation/part-486

ex:o1-per-token-models

—

achievesSpeedblah/watt-activation/part-486

1,750-9,300 tok/s

—

usesGenerationMethodblah/watt-activation/part-486

Token-by-token

—

supportsMultipleEncodingsblah/watt-activation/part-392

null

—

implementsRotorblah/watt-activation/part-392

ex:rotor-model

—

addsEncodingOptionblah/watt-activation/part-392

ex:qpsk-encoding

—

existsAsEntityblah/watt-activation/part-420

ex:null

—

hasKprobesblah/watt-activation/part-420

—

hasHyperparamLblah/watt-activation/part-420

—

hasReadoutDimblah/watt-activation/part-420

—

hasHyperparamHblah/watt-activation/part-420

—

hasHyperparamGblah/watt-activation/part-420

—

hasNumHeadsblah/watt-activation/part-420

—

hasParamsblah/watt-activation/part-420

49506

—

hasSidecarLogblah/watt-activation/part-420

logs/phase_metrics_resonant_20260319_184655.jsonl

—

referencesClaudeblah/watt-activation/part-429

ex:claude-md

—

typeblah/watt-activation/384

ex:ModelArchitecture

—

labelblah/watt-activation/384

ResonantWireLM

—

paramCountblah/watt-activation/384

33K

—

readinessStatusblah/watt-activation/384

ready to train

—

trainingRequirementblah/watt-activation/384

simple training loop

—

usesStandardTrainerblah/watt-activation/384

false

—

paramCountExactblah/watt-activation/384

33881

—