Research phase · F257

F257 — Intent Classifier

F257 covers “Intent Classifier”. The canonical record classifies it as candidate/partial, based on 3 source artifacts and 1 original source figure. The highlighted measurements are Holdout Accuracy = 0.997, Confident Accuracy = 0.9977969156819547, and Pass Rate = 0.9.

F257 Candidate/Partial Not recorded

StatusCandidate/Partial

VerdictNot recorded

Source figures1

Artifacts3

Purpose and research question

The objective is to document the recorded outcome for “Intent Classifier”. The research question is whether the phase-specific checks support the canonical candidate/partial classification. The recorded measurements include Holdout Accuracy = 0.997, Confident Accuracy = 0.9977969156819547, and Pass Rate = 0.9.

Method

F257 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

Holdout Accuracy: 0.997
Confident Accuracy: 0.9977969156819547
Pass Rate: 0.9
Classifier Probe Passed: 10
Passed: 45
Confident Total: 9986
Intent Count: 14
Adam Used: No

Recorded evidence

Holdout Accuracy	0.997
Confident Accuracy	0.9977969156819547
Pass Rate	0.9
Classifier Probe Passed	10
Passed	45
Confident Total	9986
Intent Count	14
Adam Used	No

Public evidence and source figures

F257 benchmark evidence — Intent Classifier 257 — **Intent Classifier 257**
Original source figure · 1920 × 1088 px

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

Result and interpretation

F257 Intent Classifier Not recorded

F257 retains its candidate/partial classification and the recorded decision no explicit terminal verdict. The recorded measurements include Holdout Accuracy = 0.997, Confident Accuracy = 0.9977969156819547, Pass Rate = 0.9, and Classifier Probe Passed = 10.

Key results

Holdout Accuracy: 0.997
Confident Accuracy: 0.9977969156819547
Pass Rate: 0.9
Classifier Probe Passed: 10
Passed: 45
Confident Total: 9986
Intent Count: 14
Adam Used: 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.

INTENT CLASSIFIER 257 (JSON)
INTENT CLASSIFIER 257 (MD)
INTENT CLASSIFIER CANDIDATE 257 (JSON)

Research topics

Word / World Document · Training · Evaluation