Research phase · F385

F385 — Learned Mouth Expansion

F385 covers “Learned Mouth Expansion”. The canonical record classifies it as validated, based on 4 source artifacts and 4 original source figures. The highlighted measurements are Eval Passed = 10, Passed = 4, and Eval Total = 12.

F385 Validated PASS

StatusValidated

VerdictPASS

Source figures4

Artifacts4

Purpose and research question

The objective is to document the recorded outcome for “Learned Mouth Expansion”. The research question is whether the phase-specific checks support the canonical validated classification. The recorded measurements include Eval Passed = 10, Passed = 4, and Eval Total = 12.

Method

F385 is documented by 2 structured JSON records and 2 Markdown reports and 4 original source figures. The reported table retains safe scalar measurements found in that phase package.

Documentation basis

4 source artifacts and 4 phase-specific figures are recorded for this phase.

Evidence highlights

Eval Passed: 10
Passed: 4
Eval Total: 12
Transformer Used: No
Adam Used: No
Torch Optim Used: No
Global Backward Used: No
Policy Rows Count: 5

Recorded evidence

Eval Passed	10
Passed	4
Eval Total	12
Transformer Used	No
Adam Used	No
Torch Optim Used	No
Global Backward Used	No
Policy Rows Count	5

Public evidence and source figures

F385 benchmark evidence — F385 Default Mouth Connection 386 — **F385 Default Mouth Connection 386**
Original source figure · 1800 × 900 px

F385 benchmark evidence — Learned Mouth Expansion 385 — **Learned Mouth Expansion 385**
Original source figure · 1800 × 900 px

F385 benchmark evidence — Long Chat F385 Default 387 — **Long Chat F385 Default 387**
Original source figure · 1800 × 900 px

F385 benchmark evidence — Policy Score Externalization 385 — **Policy Score Externalization 385**
Original source figure · 1800 × 900 px

These are original source figures from the existing public evidence set, not editorial HTML visualizations.

Result and interpretation

F385 Learned Mouth Expansion PASS

F385 is classified as validated with the recorded decision PASS. The recorded measurements include Eval Passed = 10, Passed = 4, Eval Total = 12, and Transformer Used = No.

Key results

Eval Passed: 10
Passed: 4
Eval Total: 12
Transformer Used: No
Adam Used: No
Torch Optim Used: No
Global Backward Used: No
Policy Rows Count: 5

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.

LEARNED MOUTH EXPANSION 385 (JSON)
LEARNED MOUTH EXPANSION 385 (MD)
POLICY SCORE EXTERNALIZATION 385 (JSON)
POLICY SCORE EXTERNALIZATION 385 (MD)

Research topics

Word Model · Word / World Document · Speech · Reasoning