Bukit Timah Tutor Secondary Mathematics

WhatsApp Us Here

NV-10 — Transfer Collapse (Template Overfit / κ-drift) (Almost-Code Canonical) v1.0

Looks competent in one format, collapses on context swap.

Summary (Canonical)

Transfer collapse happens when learning/training produces template performance instead of corridor understanding.
The system becomes overfit to a narrow pattern (κ-drift), so when the context changes, output collapses.
This creates false competence, hides decay, and reduces regeneration throughput (Ġ), accelerating civilisation drift.

1) The Root Error (Negative Void)

The absent function

Missing: transfer gates (context swap stress tests).

Most systems reward:

  • repeatable format success
  • memorisation of patterns
  • rehearsed outputs

But they don’t enforce:

  • paraphrase
  • explanation in new framing
  • novel variant problems
  • different audience / different constraints

So competence is local, not transferable.

2) Core Mechanism (κ-drift)

Let:

  • (K_{train}) = training context distribution (narrow)
  • (K_{real}) = real-world context distribution (wide)
  • κ = overfit coefficient (template dependency)

When training is narrow:
[
\kappa \uparrow
]
and transfer reliability (T\downarrow).

So under context swap:

  • retrieval fails
  • binds don’t generalise
  • corridor traversal breaks
  • output collapses

This is why “practice more of the same” can worsen the problem.

3) Observable Signs

Z0 (student)

  • scores in familiar question types
  • fails when wording changes
  • cannot explain reasoning
  • writing collapses when topic changes

Z2 (school/company)

  • staff handle standard cases
  • fail on edge cases or unusual incidents
  • escalation frequency rises
  • playbooks don’t generalise

Z4 (nation)

  • workforce can operate legacy systems
  • struggles to adopt new tech
  • brittle response to new threats
  • innovation transfer weak

4) The Transfer Collapse Corridor

  1. Training optimises one template
  2. κ-drift increases (dependence on the template)
  3. Metrics show success (Oracle failure if transfer not measured)
  4. Context changes (exam variant / new tool / shock)
  5. Performance collapses
  6. Repair workload spikes; confidence drops
  7. Regeneration throughput falls; drift accelerates

5) Hidden Fragility (Why it looks like mastery)

Template success creates:

  • fast improvement
  • stable marks in practice papers
  • confidence

But it is brittle.
True mastery shows up when:

  • question changes
  • time pressure rises
  • explanation required

Transfer is the truth test.

6) Failure Mode Trace (Required)

Template training → κ-drift increases → transfer gates absent → false competence → context swap → corridor breaks → output collapse → repair overload → regeneration throughput falls (Ġ↓) → drift risk rises.

7) Safety Conditions (Prevent NV-10)

To prevent transfer collapse:

  1. Weekly transfer tests (context swap mandatory)
  2. Explain-why requirement (binds, not answers)
  3. Variant generators (new wording, new constraints)
  4. Stress gating (T1–T3 from Idea Lattice: time, critique, transfer)
  5. Repair loops focused on bind gaps, not repetition

Almost-Code Spec Block (Copyable)

NegativeVoid.NV10.TransferCollapse.KappaDrift.v1.0

Negative Void:
  Training optimises template success; lacks transfer gates
Model:
  κ := overfit coefficient (template dependency)
  Narrow K_train -> κ increases -> transfer reliability T decreases
Observable Signs:
  good in familiar format, collapses on variants; cannot explain reasoning
Failure Mode Trace:
  template training -> κ↑ -> no transfer gates -> false competence ->
  context swap -> corridor breaks -> output collapse -> repair overload ->
  Ġ↓ -> drift risk↑
Safety Conditions:
  weekly transfer tests + explain-why + variant generators + stress gating + targeted repair

FAQ (Short)

Q1: Is using templates always bad?
No. Templates are useful as scaffolds. The failure is stopping there and never testing transfer.

Q2: What’s the simplest transfer gate?
Same concept, different wording + different example + timed explanation.

Q3: Why does this matter for civilisation?
A civilisation overfit to legacy templates fails when environments change—its regeneration can’t keep up with new load.

Start Here: 

Start Here:

eduKateSG Learning Systems: 

Recommended Internal Links (Spine)

Start Here for Lattice Infrastructure Connectors

Start here if you want the full sequence:

Vocabulary OS Series Index:

Fence English Learning System: 

eduKateSG Learning Systems: 

Recommended Internal Links (Spine)

Start Here for Lattice Infrastructure Connectors