Dontopedia

Current Model

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-11.)

Current Model has 35 facts recorded in Dontopedia across 16 references, with 1 live disagreement.

35 facts·33 predicates·16 sources·1 in dispute

Mostly:rdf:type(3), is a(1), learns by imitation(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (15)

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.

aboutAbout(2)

believesMinimalSizeReachedBelieves Minimal Size Reached(1)

causesConfusionIfMixedCauses Confusion If Mixed(1)

comparesPerplexityWithCompares Perplexity With(1)

demonstratesEarlyTrainingDemonstrates Early Training(1)

developModelDevelop Model(1)

existInExist in(1)

generatedByGenerated by(1)

instillsQaPatternInstills Qa Pattern(1)

isBeingTrainedOnIs Being Trained on(1)

largerThanLarger Than(1)

referencesExistingWorkReferences Existing Work(1)

refersToModelRefers to Model(1)

wouldDoGreatWould Do Great(1)

Other facts (35)

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.

35 facts
PredicateValueRef
Rdf:typeStatistical Model[12]
Rdf:typeModel[14]
Rdf:typeModel[15]
Is aClone Model[1]
Learns by ImitationOnly[2]
Is145m Parameters145000000[3]
Has Max Seq Len2048[4]
Under Trainingtrue[4]
Trained on2k Windows[4]
Needs Improvementtrue[5]
Currently Producesessay fragments[5]
Has Effective Params37M[5]
Improves Overessay fragments now[5]
Would Benefit From Renamingconsistent behavior[6]
Is Needed BaselineTrue[7]
Existstrue[8]
Lacks FeatureMemory Mapping[9]
Should HaveFused Kernel[10]
Thinks in Scale Count~1[11]
Compared Favorably toAnchor Kan 65k Run[13]
Uses Flops50[13]
Flops Comparison BasisAnchor Kan 65k Run[13]
Optimized byModified Adam[13]
Uses Geometric Structure for AttentionN Dim Sphere[13]
Has Nats Per Byte1.38[15]
Has Bpb1.99[15]
Has Iteration Count6000[15]
Has Prediction Quality Ratio4.7[15]
Has Parameter Count8400000[15]
Compares With Parameter Factor15× fewer parameters[15]
Operates onraw bytes[15]
Has Absence oftokenizer[15]
Compared Against Categoryother byte-level models at similar scale[15]
IsBert Base Uncased[16]
May BeSuboptimal[16]

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.

35 undated
—
isAblah/tpmjs/part-20
ex:clone-model
—
learnsByImitationblah/vidya/part-7
ex:only
—
is145m-parametersblah/watt-activation/part-97
145000000
—
hasMaxSeqLenblah/watt-activation/part-126
2048
—
underTrainingblah/watt-activation/part-126
true
—
trainedOnblah/watt-activation/part-126
ex:2k-windows
—
needsImprovementblah/watt-activation/part-163
true
—
currentlyProducesblah/watt-activation/part-163
essay fragments
—
hasEffectiveParamsblah/watt-activation/part-163
37M
—
improvesOverblah/watt-activation/part-163
essay fragments now
—
wouldBenefitFromRenamingblah/watt-activation/part-217
consistent behavior
—
isNeededBaselineblah/watt-activation/part-271
ex:true
—
existsblah/watt-activation/part-467
true
—
lacksFeatureblah/watt-activation/part-663
ex:memory-mapping
—
shouldHaveblah/watt-activation/part-677
ex:fused-kernel
—
thinksInScaleCountblah/watt-activation/part-354
~1
—
typebeam/ddefc08a-c24b-460a-9fa2-07d14a817398
ex:StatisticalModel
—
comparedFavorablyToblah/watt-activation/153
ex:AnchorKAN-65k-run
—
usesFlopsblah/watt-activation/153
50
—
flopsComparisonBasisblah/watt-activation/153
ex:AnchorKAN-65k-run
—
optimizedByblah/watt-activation/153
ex:modified-adam
—
usesGeometricStructureForAttentionblah/watt-activation/153
ex:n-dim-sphere
—
typeblah/watt-activation/275
ex:Model
—
typeblah/watt-activation/336
ex:Model
—
hasNatsPerByteblah/watt-activation/336
1.38
—
hasBPBblah/watt-activation/336
1.99
—
hasIterationCountblah/watt-activation/336
6000
—
hasPredictionQualityRatioblah/watt-activation/336
4.7
—
hasParameterCountblah/watt-activation/336
8400000
—
comparesWithParameterFactorblah/watt-activation/336
15× fewer parameters
—
operatesOnblah/watt-activation/336
raw bytes
—
hasAbsenceOfblah/watt-activation/336
tokenizer
—
comparedAgainstCategoryblah/watt-activation/336
other byte-level models at similar scale
—
isbeam/63f3f6ff-b059-492e-954d-ccca67c2349d
ex:bert-base-uncased
—
mayBebeam/63f3f6ff-b059-492e-954d-ccca67c2349d
ex:suboptimal

References (16)

16 references
  1. [1]Part 201 fact
    ctx:discord/blah/tpmjs/part-20
  2. [2]Part 71 fact
    ctx:discord/blah/vidya/part-7
  3. [3]Part 971 fact
    ctx:discord/blah/watt-activation/part-97
  4. [4]Part 1263 facts
    ctx:discord/blah/watt-activation/part-126
  5. [5]Part 1634 facts
    ctx:discord/blah/watt-activation/part-163
  6. [6]Part 2171 fact
    ctx:discord/blah/watt-activation/part-217
  7. [7]Part 2711 fact
    ctx:discord/blah/watt-activation/part-271
  8. [8]Part 4671 fact
    ctx:discord/blah/watt-activation/part-467
  9. [9]Part 6631 fact
    ctx:discord/blah/watt-activation/part-663
  10. [10]Part 6771 fact
    ctx:discord/blah/watt-activation/part-677
  11. [11]Part 3541 fact
    ctx:discord/blah/watt-activation/part-354
  12. [12]Ddefc08a C24b 460a 9fa2 07d14a8173981 fact
    ctx:claims/beam/ddefc08a-c24b-460a-9fa2-07d14a817398
  13. [13]1535 facts
    ctx:discord/blah/watt-activation/153
    • full textwatt-activation-153
      text/plain3 KBdoc:agent/watt-activation-153/adeb3de4-7496-4fe9-a51b-b54f859c8ef4
      Show excerpt
      [2026-03-09 16:00] xenonfun: step 2100/16684 12.6% loss=6.0128 ppl= 408.6 lr=4.87e-05 668ms 12,271tok/s eta=162min step 2200/16684 13.2% loss=6.0229 ppl= 412.8 lr=4.85e-05 665ms 12,326tok/s eta=160min step 2300/16684 …
  14. [14]2751 fact
    ctx:discord/blah/watt-activation/275
    • full textwatt-activation-275
      text/plain3 KBdoc:agent/watt-activation-275/149bf24b-54fb-4412-b289-e2e03bccffe2
      Show excerpt
      [2026-03-13 20:46] xenonfun: ⏺ This tells a clear story: Current trained weights are NOT naturally ternary. Only ~46% of values are near {-1, 0, 1} after scaling (random Gaussian would give ~47%, so no better than chance). Relative err…
  15. [15]33610 facts
    ctx:discord/blah/watt-activation/336
    • full textwatt-activation-336
      text/plain3 KBdoc:agent/watt-activation-336/04f318bf-4029-460c-b2ce-82900263e51e
      Show excerpt
      [2026-03-15 15:12] xenonfun: ⏺ Step 2000 results (bs=512 seq=256 (its pointless to use higher bandwidth cuts off hurts quality of mappings beyond this)) so trying optimal run, high BS smooth out variance considerable. Eval (mixed_bytes v…
  16. [16]63f3f6ff B059 492e 954d Ccca67c2349d2 facts
    ctx:claims/beam/63f3f6ff-b059-492e-954d-ccca67c2349d
    • full textbeam-chunk
      text/plain1020 Bdoc:beam/63f3f6ff-b059-492e-954d-ccca67c2349d
      Show excerpt
      However, I'm only achieving about 80% accuracy with this approach. I've studied LLM-based reformulation and noted a 25% intent accuracy boost for 6,000 complex queries. Can you help me improve my implementation to reach at least 92% detecti…

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.

↑