MAOE: Mixture of Anticipatory Orthogonal Experts
MAOE separates experts by authority as well as specialization. The architecture is meant for recognition stacks where some modules may propose text, some may score acoustic evidence, and only governed combinations are allowed to become accepted outputs.
Paper workspace
Live draft structure
Artifacts
Architecture source
MAOE is mapped as an architecture, not as a trained ASR result.
source-only
Editable source
Technical architecture source exists. Empirical MAOE-vs-baseline proof is still pending.
Source anchors
MAOE-NKo-Technical-Architecture.md
LUME-CC/07-nko/maoe-routing.md
NKO-CONSOLIDATION.md
Method tags
Ingest intersections
Status
Architecture materialized; empirics pending.
Key claims
01
Expert routing should include admissibility, not only confidence.
02
Different modules should have different authority to assert truth.
03
The N'Ko ASR stack needs governed composition before self-improvement.
Public reading note
Architecture page is safe; empirical claims remain pending.
Standard skeleton
What this paper must keep proving
problem
Post-ASR correction becomes unsafe when every expert is allowed to rewrite the transcript with equal authority.
method
Route chunks into stable, boundary, uncertain, recovery, and novelty partitions, each with a different correction authority.
implementation
Technical architecture around the verified CTC anchor, AGP proposal lane, TurboQuant retrieval, and Rust gate.
data
Partition-specific correction rows are required before MAOE can become a trained correction expert.
evaluation
Head-to-head against the verified ASR checkpoint on the same snapshot, with accept/reject and CER deltas.
references
Mixture-of-experts routing, ASR rescoring, retrieval-augmented correction, safety/admissibility gates.
openQuestions
The full MAOE bridge-row training wave and same-snapshot evaluation remain pending.
Checkpoints and references
Proof chain
Claim checkpoint
central-claim slot
Every central claim must point to a proof anchor or remain labeled as speculative.
Implementation checkpoint
implementation-map slot
Every method should identify the code path, harness, schema, or protocol that embodies it.
Evidence checkpoint
evidence-manifest slot
Every reported result should point to run IDs, packet IDs, data snapshots, commits, or review artifacts.
Reference checkpoint
references slot
Every external claim should resolve to a cited paper, benchmark, standard, or documented prior system.
Release checkpoint
release-gate slot
Every PDF needs a named condition before it can move from draft to citation-ready.