MAOE Routing
MAOE means Mixture of Anticipatory Orthogonal Experts. In this project it should be documented as a routing/correction pattern, not as the verified ASR acoustic anchor.
Full Public Reader
MAOE Routing
MAOE means Mixture of Anticipatory Orthogonal Experts. In this project it should
be documented as a routing/correction pattern, not as the verified ASR acoustic
anchor.
Verified Correction
The verified N'Ko ASR anchor is trajectory-biased Transformer CTC:
- `Desktop/nko-brain-scanner/asr/trajectory_asr.py`
- `Desktop/nko-brain-scanner/constrained/trajectory_bias.py`
MAOE-style work appears around ASR output:
- `Desktop/Comp-Core/experiments/agp_mlx/asr_bridge/partition_policy.py`
- `Desktop/Comp-Core/experiments/agp_mlx/asr_bridge/expert_router.py`
So the correct chain is:
audio -> trajectory-biased CTC -> ASR hypothesis/telemetry
-> MAOE-style partition/routing -> preserve/complete/repair/recover/blockWhy MAOE Still Matters
MAOE is important because N'Ko and embodied motion both need authority control.
When evidence is stable, preserve it. When evidence is uncertain, limit repairs.
When evidence is novel, prevent a generic model from normalizing it away.
That same principle can later help motion:
camera/mocopi/body evidence
-> gesture hypothesis
-> authority lane
-> command or no commandBut that is a future design pattern unless implemented in the movement code.
Movement-Side Status
Do not say "movement uses MAOE" unless source proves it.
Current source-grounded terms are:
- EchelonCore;
- LatentUpdater;
- SANPipeline;
- BodyTruth;
- AirDeck safety;
- ClaimBridge.
MAOE can be proposed for future gesture ambiguity handling, especially for
AirDeck commands where false positives are costly.
Orthogonality
Orthogonality is a useful design constraint for keeping experts or authority
lanes from collapsing into the same behavior. It is not automatically present in
every MoE or routing implementation.
If movement adopts MAOE later, the docs should specify:
- what the experts are;
- what their authority lanes mean;
- what data trains them;
- what metric proves improvement;
- how K11 safety vetoes commands.
Source-Grounded Summary
Use:
> MAOE is a correction/admissibility routing pattern around the N'Ko ASR anchor
> and a plausible future pattern for movement/gesture routing.
Avoid:
> MAOE trained the ASR baseline.
> Movement already uses MAOE.
> MAOE and the motion latent are the same architecture.
Promotion Decision
Attach run IDs, datasets, metrics, and reproduction commands.
Source Anchor
computational-choreography/07-nko-synthesis/maoe-routing.md
Detected Structure
Method · Evaluation · Code Anchors · Architecture