Skip to content

Is the ~22% linearization ceiling a capacity wall, or a recipe limit? #1

Description

@ecloud-bh

What we measured

Co-trained, spread configurations on TurkishMMLU (oracle vs hybrid, identical protocol, paired bootstrap 95% CI):

Linear MMLU Δ 95% CI Verdict
20% (8 layers) −0.7 [−2.9, +1.7] not significant
22.5% (9 layers) −1.8 [−4.2, +0.7] not significant (limited power)
25% (10 layers) −4.0 [−6.4, −1.4] significant degradation

Our diagnosis (input teacher-forcing + cosine analysis) attributes the loss roughly half to cross-layer compounding and half to a single-layer capacity gap, which is consistent with a genuine capacity wall in the middle band.

The open question

Is 25% failing because those layers genuinely cannot be represented by a fixed-size state — or because our co-training recipe is too thin? We used KL distillation on the GDN parameters only. We did not try:

  • MLP-LoRA on the downstream feed-forward blocks
  • hidden-state matching in addition to logit KL
  • substantially more co-training data or steps
  • multiple seeds (all main comparisons are single-seed)

In one earlier experiment a fuller recipe closed roughly 40% of the gap (−9.3 → −5.3) and then plateaued — suggesting the wall is real but that our current number may be pessimistic.

What would settle it

Co-train the 25% configuration with an expanded recipe and report MMLU with a paired bootstrap CI on the held-out set. Even a partial result is useful. If the gap closes below significance, the ceiling is a recipe artifact; if it plateaus again, the capacity interpretation stands.

Protocol and raw predictions: REPRODUCE.md, results/raw_predictions.json.

Metadata

Metadata

Assignees

No one assigned

    Labels

    help wantedExtra attention is neededmeasurementBir ölçüm gerektirirresearch questionAçık araştırma sorusu — ölçüm bekliyor

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions