Dontopedia

Our Model

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-05-02.)

Our Model has 26 facts recorded in Dontopedia across 3 references.

26 facts·26 predicates·3 sources

Mostly:equivalent to(1), achieved bpb drop(1), has parameters(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (4)

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.

has14xMoreParametersHas14x More Parameters(1)

has15xMoreTrainingStepsHas15x More Training Steps(1)

hasZeroCostHas Zero Cost(1)

zeroComputationalCostZero Computational Cost(1)

Other facts (26)

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.

26 facts
PredicateValueRef
Equivalent toLoheffn V2[1]
Achieved Bpb Drop2.39 to 1.572[2]
Has Parameters8.25M[2]
Is3x Better ThanVq Model[2]
Is Still Improvingtrue[2]
Is Within1 6x ofGpt 2[2]
Nears Gpt2 Performance1.6x worse[2]
Outperforms Ngram LevelBpb 1 572[2]
Predicted to Improve FurtherFuture Training Runs[2]
Presupposes Char LevelGood Char Level Model[2]
Shares Training Budget20K steps[2]
Crossed Territory FromBasic Statistics Territory[2]
Crossed Territory toLegitimate Neural Language Model Territory[2]
Entered Legitimate Neural Lm Territorytrue[2]
Has Bpb Score1.572[2]
Has CheckpointsBest Checkpoints[2]
Has Improvement Slope-0.052 BPB per 1K steps[2]
Byte Leveltrue[3]
Smaller Scale Advantagetrue[3]
Has Params8400000[3]
Operates on Raw Bytestrue[3]
At Iterations6000[3]
Has Bpb1.99[3]
Progressing Welltrue[3]
Worse Than Gpt2 by Factor4.7[3]
Nats Per Byte1.38[3]

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.

equivalentToblah/watt-activation/part-160
ex:loheffn-v2
achievedBpbDropblah/watt-activation/part-326
2.39 to 1.572
hasParametersblah/watt-activation/part-326
8.25M
is3xBetterThanblah/watt-activation/part-326
ex:vq-model
isStillImprovingblah/watt-activation/part-326
true
isWithin1-6xOfblah/watt-activation/part-326
ex:gpt-2
nearsGpt2Performanceblah/watt-activation/part-326
1.6x worse
outperformsNgramLevelblah/watt-activation/part-326
ex:bpb-1-572
predictedToImproveFurtherblah/watt-activation/part-326
ex:future-training-runs
presupposesCharLevelblah/watt-activation/part-326
ex:good-char-level-model
sharesTrainingBudgetblah/watt-activation/part-326
20K steps
crossedTerritoryFromblah/watt-activation/part-326
ex:basic-statistics-territory
crossedTerritoryToblah/watt-activation/part-326
ex:legitimate-neural-language-model-territory
enteredLegitimateNeuralLmTerritoryblah/watt-activation/part-326
true
hasBpbScoreblah/watt-activation/part-326
1.572
hasCheckpointsblah/watt-activation/part-326
ex:best-checkpoints
hasImprovementSlopeblah/watt-activation/part-326
-0.052 BPB per 1K steps
byteLevelblah/watt-activation/part-338
true
smallerScaleAdvantageblah/watt-activation/part-338
true
hasParamsblah/watt-activation/part-338
8400000
operatesOnRawBytesblah/watt-activation/part-338
true
atIterationsblah/watt-activation/part-338
6000
hasBpbblah/watt-activation/part-338
1.99
progressingWellblah/watt-activation/part-338
true
worseThanGpt2ByFactorblah/watt-activation/part-338
4.7
natsPerByteblah/watt-activation/part-338
1.38

References (3)

3 references
  1. [1]Part 1601 fact
    ctx:discord/blah/watt-activation/part-160
  2. [2]Part 32616 facts
    ctx:discord/blah/watt-activation/part-326
  3. [3]Part 3389 facts
    ctx:discord/blah/watt-activation/part-338

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.