Dontopedia

Language Model Without Tokenizer

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

Language Model Without Tokenizer has 7 facts recorded in Dontopedia across 2 references.

7 facts·7 predicates·2 sources

Mostly:lacks tokenizer(1), encodes(1), avoids symbol formation(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (2)

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.

builtBuilt(1)

discussesTopicDiscusses Topic(1)

Other facts (7)

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.

7 facts
PredicateValueRef
Lacks TokenizerTokenizer[1]
EncodesRaw Bytes[1]
Avoids Symbol Formationnull[1]
Uses EncodingPhase Modulated Oscillators[1]
Rdf:typeModel Architecture[2]
Lacks FeatureTokenizer[2]
Uses ApproachRaw Byte Encoding[2]

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.

lacksTokenizerblah/watt-activation/part-325
ex:tokenizer
encodesblah/watt-activation/part-325
ex:raw-bytes
avoidsSymbolFormationblah/watt-activation/part-325
null
usesEncodingblah/watt-activation/part-325
ex:phase-modulated-oscillators
typeblah/watt-activation/323
ex:ModelArchitecture
lacksFeatureblah/watt-activation/323
ex:tokenizer
usesApproachblah/watt-activation/323
ex:raw-byte-encoding

References (2)

2 references
  1. [1]Part 3254 facts
    ctx:discord/blah/watt-activation/part-325
  2. [2]3233 facts
    ctx:discord/blah/watt-activation/323
    • full textwatt-activation-323
      text/plain3 KBdoc:agent/watt-activation-323/60722a8e-0dbe-4aee-be7b-862e43a25b6f
      Show excerpt
      [2026-03-15 03:41] xenonfun: ``` Elevator Pitch We built a language model that doesn't use a tokenizer. Instead of BPE or word pieces, we encode raw bytes directly as phase-modulated oscillators using digital communications theory

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.