F706-D — Sequential Composition Repair
F706-D covers “Sequential Composition Repair”. The canonical record classifies it as candidate/partial, based on 3 source artifacts and 1 original source figure. The highlighted measurements are Decision = PARTIAL_EFFECT_NOT_ENOUGH, Gates Open Gain Gte 15pp = No, and Gates Known Accuracy Gte 30pct = No.
Purpose and research question
The objective is to document the recorded outcome for “Sequential Composition Repair”. The research question is whether the phase-specific checks support the canonical candidate/partial classification. The recorded measurements include Decision = PARTIAL_EFFECT_NOT_ENOUGH, Gates Open Gain Gte 15pp = No, and Gates Known Accuracy Gte 30pct = No.
Method
F706-D is documented by 2 structured JSON records and 1 Markdown report and 1 original source figure. The reported table retains safe scalar measurements found in that phase package.
Documentation basis
3 source artifacts and 1 phase-specific figure are recorded for this phase.
Evidence highlights
- Decision: PARTIAL_EFFECT_NOT_ENOUGH
- Gates Open Gain Gte 15pp: No
- Gates Known Accuracy Gte 30pct: No
- Gates Token Recall Gte 50pct: No
- Gates Slot Completion Gte 60pct: Yes
- Gates Subject Retention Gte 80pct: No
- Gates Causal Completion Gte 60pct: No
- Gates Premature Stop Lte 40pct: No
Recorded evidence
| Decision | PARTIAL_EFFECT_NOT_ENOUGH |
|---|---|
| Gates Open Gain Gte 15pp | No |
| Gates Known Accuracy Gte 30pct | No |
| Gates Token Recall Gte 50pct | No |
| Gates Slot Completion Gte 60pct | Yes |
| Gates Subject Retention Gte 80pct | No |
| Gates Causal Completion Gte 60pct | No |
| Gates Premature Stop Lte 40pct | No |
Public evidence and source figures
Original source figure · 2550 × 1020 px
These are original source figures from the existing public evidence set, not editorial HTML visualizations.
Result and interpretation
F706-D retains its candidate/partial classification and the recorded decision no explicit terminal verdict. The recorded measurements include Decision = PARTIAL_EFFECT_NOT_ENOUGH, Gates Open Gain Gte 15pp = No, Gates Known Accuracy Gte 30pct = No, and Gates Token Recall Gte 50pct = No.
Key results
- Decision
- PARTIAL_EFFECT_NOT_ENOUGH
- Gates Open Gain Gte 15pp
- No
- Gates Known Accuracy Gte 30pct
- No
- Gates Token Recall Gte 50pct
- No
- Gates Slot Completion Gte 60pct
- Yes
- Gates Subject Retention Gte 80pct
- No
- Gates Causal Completion Gte 60pct
- No
- Gates Premature Stop Lte 40pct
- No
Evidence sources
Values on this page are reproduced from the archived phase record and the source artifacts below. No new benchmark was run to create this page.
- SEQUENTIAL COMPOSITION BLIND FREEZE 706D (JSON)
- SEQUENTIAL COMPOSITION REPAIR 706D (JSON)
- SEQUENTIAL COMPOSITION REPAIR 706D (MD)