Baseline mean
Token-pair enumeration for one complete 65,536-token plan construction.
Measured evidence
Exact sparse-plan construction
Exact plan-identity checks and host-local construction timing for one sparse-attention planning component.
Recorded finding: Across 20 measurements per arm, direct block-region counting retained the tested plan identity and reduced mean 65,536-token plan-construction time from 784.49 ms to 7.99 ms on the disclosed host.
On this page
Currency and next evidence gate
Decision signals
Key metrics and controls
The recorded values, with their controls and scope.
Candidate mean
Exact block-region counting on the same host and plan protocol.
Component wall-time reduction
Component-only engineering result; not end-to-end performance.
Tested plan identity
20 timing measurements per arm; identity checked separately.
Source-linked visual evidence
What the evidence shows
Open any chart’s data table to inspect the exact values, or download the CSV.
Source-linked comparison
Mean sparse-plan construction time
Host-local time to construct one complete 65,536-token sparse-attention plan before and after exact block-region counting.
Finding: The candidate retained the tested plan identity while reducing this component's mean wall time from 784.49 ms to 7.99 ms.
Token-pair enumeration
Block-region counting
View source values
Scroll sideways if any columns are out of view.
| Group | Mean wall time |
|---|---|
| Token-pair enumeration | 784.49 ms |
| Block-region counting | 7.99 ms |
Method
How to read this result
- The measured component is sparse-attention plan construction, not attention execution, training, or generation.
- Twenty complete plan constructions were measured per arm on the same disclosed CPU host at a sequence length of 65,536 tokens.
- The baseline enumerated causal token pairs; the candidate counted the same selected block regions directly.
- Plan hashes, pair counts, partial-final-block cases, and focused regression tests were used to check identity.
What this result does not establish
This is component-level engineering evidence. It does not establish end-to-end speed, model quality, memory or energy savings, cross-host performance, or superiority.
- End-to-end training or generation speedup
- Model quality or generalisation
- Memory or energy savings
- Cross-host performance
- A public superiority claim
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 JSONProvenance
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 receiptContinue reviewing
Related evidence
measuredbounded
Historical- Observed
- 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.
- Supports
- Real Qwen3 0.6B fracture, exact K32 engineering reconstruction, and a completed protected four-way trace for material-sparsity alignment.
- Does not support
- 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.
- Next decisive evidence
- Use the retained negative geometry to test refracture and recovery, then rerun the material-sparsity ladder under quality-bearing controls.
Decision relevance: research · investment · diligence
Review evidencecontrolreviewed
Superseded- 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