NeuroForge
Appearance

Use your device setting or keep a preference on this browser.

Contact
Navigate
Support the labTalk to Lloyd

Quick find

Find what you need

Esc

Try a job or topic, such as meeting notes.

    Search learning resources Free lessons, printables and activities.

    Measured evidence

    Model fracture milestone

    Real Qwen3 0.6B fracture, exact K32 engineering reconstruction, and a completed protected four-way trace for material-sparsity alignment.

    Recorded finding: All 28 donor layers produced balanced expert assignments, K32 matched five frozen engineering cases exactly, and all 112 protected trace units completed across four isolated roles.

    On this page

    Currency and next evidence gate

    Decision signals

    Key metrics and controls

    The recorded values, with their controls and scope.

    28/28

    Donor layers traced

    32 balanced experts per layer; 96 neurons each.

    5/5

    Exact frozen cases

    First logits, first-token top-five order, and complete greedy generations at K32.

    0.0

    Maximum first-logit error

    Observed across the five registered engineering cases; broader token-level comparison remains a next gate.

    112/112

    Protected trace units

    4 isolated roles across all 28 layers; no failed or interrupted unit was promoted.

    Source-linked visual evidence

    What the evidence shows

    Open any chart’s data table to inspect the exact values, or download the CSV.

    Status matrix

    Exact reconstruction checks

    Matched and compared case counts for the three frozen K32 engineering-parity checks.

    Finding: All three checks matched in all five cases; this is exact observed reconstruction within the frozen set, not broad quality evidence.

    Direct linkDownload CSV
    Exact reconstruction checksMatched and compared case counts for the three frozen K32 engineering-parity checks. All three checks matched in all five cases; this is exact observed reconstruction within the frozen set, not broad quality evidence.0246casesMatchedComparedFirst-token logitsFirst-token logits55First-token top fiveFirst-token top five55Greedy generationsGreedy generations55
    View source values

    Scroll sideways if any columns are out of view.

    Exact reconstruction checks; values in cases
    GroupMatchedCompared
    First-token logits5 cases5 cases
    First-token top five5 cases5 cases
    Greedy generations5 cases5 cases

    Ordered source series

    Assignment structure across 28 donor layers

    Minimum, median, and maximum difference between balanced activation-derived assignments and contiguous neuron slices.

    Finding: Every layer produced a distinct balanced assignment 96.5–97.6% different from contiguous slicing; this establishes non-arbitrary organisation, not semantic superiority.

    Direct linkDownload CSV
    Assignment structure across 28 donor layersMinimum, median, and maximum difference between balanced activation-derived assignments and contiguous neuron slices. Connecting lines show ordered observations and do not imply unmeasured intermediate results. Every layer produced a distinct balanced assignment 96.5–97.6% different from contiguous slicing; this establishes non-arbitrary organisation, not semantic superiority.050100150% differentAssignment differenceMinimum observed: Assignment difference, 96.5 % differentMedian observed: Assignment difference, 96.9 % differentMaximum observed: Assignment difference, 97.6 % differentMinimum observedMinimum observedMedian observedMedian observedMaximum observedMaximum observed

    Connecting strokes aid comparison between published ordered points; they do not represent measurements between those points.

    View source values

    Scroll sideways if any columns are out of view.

    Assignment structure across 28 donor layers; values in % different
    PointAssignment difference
    Minimum observed96.5 % different
    Median observed96.9 % different
    Maximum observed97.6 % different

    Progress and control

    Native adapter update scope

    Trainable and frozen parameter shares in the 671.315M-parameter one-step bootstrap.

    Finding: The update covered 577 injected modules while training 1.552% of parameters; it proves local numerical execution, not convergence or generation quality.

    Direct linkDownload CSV
    Native adapter update scopeTrainable and frozen parameter shares in the 671.315M-parameter one-step bootstrap. The update covered 577 injected modules while training 1.552% of parameters; it proves local numerical execution, not convergence or generation quality.050100150% of parametersTrainable adapterFrozen parametersTrainable adapterTrainable adapter1.552Frozen parametersFrozen parameters98.448
    View source values

    Scroll sideways if any columns are out of view.

    Native adapter update scope; values in % of parameters
    GroupParameter share
    Trainable adapter1.552 % of parameters
    Frozen parameters98.448 % of parameters

    Method

    How to read this result

    1. A frozen real-model activation trace covered all 28 feed-forward layers and assigned every source neuron exactly once into equal-width experts.
    2. The complete K32 expert set was reconstructed and compared with the donor on five pre-registered engineering cases; a separate native checkpoint then completed one bounded adapter step.
    3. A separate protected trace completed all 112 layer-role units across train, calibration, validation, and final-holdout roles, with distinct split identities and no fit access to the final holdout.
    4. The bounded alignment execution path is implemented and its protocol is executable; this pack does not present a complete real aligned checkpoint or terminal alignment receipt.
    5. The machine-readable source JSON retains all 28 layer rows; the visual chart presents their minimum, median, and maximum for legibility.
    Read the complete evidence method

    What this result establishes

    The bounded alignment path is executable, but no complete aligned checkpoint is promoted; K8/K16/K24 quality, savings, broader generalisation, and production readiness remain unevaluated.

    • Exact parity is limited to five frozen engineering cases at complete K32 reconstruction.
    • The activation curriculum was a small synthetic engineering set and does not establish semantic or routing superiority.
    • The adapter result is one finite update, not convergence, benchmark quality, or sustained training evidence.
    • The protected four-way trace is complete and the bounded alignment path is executable, but no complete real 28-layer aligned checkpoint or terminal alignment receipt is promoted here.
    • K8, K16, K24, and adaptive-K quality, latency, throughput, memory, and energy remain unevaluated on this converted model.
    • No production-readiness, cross-model portability, continual-learning, or capability-marketplace claim is supported by this pack.

    Inspection and reuse

    Downloads and provenance

    Source data

    Machine-readable summary

    The recorded JSON artifact is staged from the governed public evidence pack during the build. Its currency notice determines whether it may be read as current.

    Open source JSON
    Exact reconstruction checks CSVAssignment structure across 28 donor layers CSVNative adapter update scope CSV

    Provenance

    Build and evidence receipt

    Use the public build receipt to compare the deployed site source, evidence digest, claim posture, and inference release state.

    Open build receipt

    Continue reviewing

    Related evidence

    measuredreviewed
    Refresh required

    Training probe

    Observed
    Small controlled probes provide measurement signals suitable for engineering decisions and follow-up experiments.
    Supports
    Step counts, throughput ranges, parameter footprint, and pair-work reduction from bounded probes.
    Does not support
    Probe-scale behavior is not a substitute for matched end-to-end production benchmarks.
    Next decisive evidence
    Regenerate from the repaired exact artifact identity and close or retain the phase-41 contract-discovery failure.

    Decision relevance: research · engineering

    Review evidence
    controlreviewed
    Superseded

    Quality gates

    Observed
    Critical system surfaces are governed by explicit pass/fail controls rather than narrative assurance alone.
    Supports
    Configuration validation, critical gate coverage, phase-control checks, and API observability posture.
    Does not support
    Control coverage describes engineering discipline; it does not independently prove model capability.
    Next decisive evidence
    Rerun and close—or retain—the three failing raw-discovery validations, then publish an exact-SHA quality-gate pack with executed assembly CI.

    Decision relevance: governance · diligence

    Review evidence