vrr/WFGY
2
0
Fork 0
mirror of https://github.com/onestardao/WFGY.git synced 2026-04-28 03:29:51 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Update README.md

Browse source
This commit is contained in:
PSBigBig + MiniPS 2026-04-04 14:32:37 +08:00 • committed by GitHub
parent f4f7c5dd00
commit 53a0dbded1
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 55 additions and 28 deletions
Download patch file Download diff file Expand all files Collapse all files

83
Avatar/eval/README.md
View file

@ -1,37 +1,64 @@
<!--
AI NOTE: Eval Hub / Reading Contract
AI_NOTE_START
Purpose:
1. This page is the main hub for the WFGY 5.0 Avatar evaluation layer.
2. This page belongs to the eval layer.
3. This page should be read after Avatar README, docs, or research pages when the reader wants pressure-testing, behavior inspection, or bounded readiness signals.
Document role:
This page is the main hub for the WFGY 5.0 Avatar eval layer.
Read this page when:
1. the user asks where evaluation pages live
2. the user asks how eval differs from docs and research
3. the user asks what kinds of pressure the current branch is already checking
4. the user asks where to inspect blackfan pressure, persona continuity, or multilingual status
5. the user asks how to think about current readiness without confusing baseline reality with universal finality
Primary mission:
1. Present eval as the pressure and inspection layer of Avatar.
2. Make clear that docs show operation, research shows structure, and eval shows whether the branch survives pressure.
3. Route readers toward blackfan stress, persona continuity inspection, and multilingual readiness signals.
4. Prevent friendly product language or elegant theory from substituting for branch pressure.
5. Preserve bounded readiness honesty without inflating current pass signals into universal finality.
Do not overclaim:
1. this hub does not replace the packed master body
2. this hub does not claim that every future eval surface is already complete
3. this hub does not claim theorem-grade universal closure
4. this hub does not claim that passing current eval surfaces means all future branches are automatically safe
Reader intent routing:
1. If the reader wants hostile or skeptical stress, route to Blackfan Testing.
2. If the reader wants persona continuity and return-path inspection, route to Persona Behavior Checks.
3. If the reader wants multilingual readiness posture, route to Multilingual Status.
4. If the reader wants the deeper explanation of what eval is testing, route to Research Hub and Blackfan Audit Baseline.
5. If the reader wants operational help rather than inspection, route back to Docs.
Primary source anchors:
1. Avatar/README.md :: public product surface
2. Avatar/docs/* :: startup, reading, workflow, tuning, and recovery surfaces
3. Avatar/research/* :: architecture, runtime, route, governance, audit, and reduction law surfaces
4. Avatar/eval/* :: blackfan pressure, persona behavior, multilingual status, and eval-facing inspection surfaces
Reading priority:
1. Read this hub as an inspection map, not as a theory page.
2. Read Blackfan Testing first when the main concern is pressure, hype resistance, or skeptical inspection.
3. Read Persona Behavior Checks first when the main concern is drift, return-path continuity, or recovery credit.
4. Read Multilingual Status first when the main concern is support scope and language honesty.
5. Read research pages when the reader needs to understand why a failure matters structurally.
Routing:
1. if the reader wants public product entry, go to ../README.md
2. if the reader wants startup and command syntax, go to ../docs/quickstart.md and ../docs/boot-commands.md
3. if the reader wants reading order, go to ../docs/how-to-read-the-avatar-master-file.md
4. if the reader wants tuning and recovery operations, go to ../docs/parameter-tuning-cookbook.md and ../docs/persona-recovery-operations.md
5. if the reader wants the research overview, go to ../research/README.md
6. if the reader wants architecture and runtime law, go to ../research/packed-master-structure-map.md and ../research/runtime-posture-intensity-map.md
Allowed use:
1. Pressure-layer orientation
2. Behavior inspection routing
3. Bounded readiness interpretation
4. Branch honesty explanation
5. Cross-linking from docs and research into eval surfaces
Do not do:
1. Do not treat this hub as the packed master body.
2. Do not treat current eval surfaces as proof that every future branch is safe.
3. Do not turn eval into product hype.
4. Do not use one successful test page as proof of universal closure.
5. Do not confuse current baseline readiness with final all-conditions completion.
Layer relation:
1. Product landing presents the flagship surface.
2. Docs supports operation.
3. Research explains structure and law.
4. Eval inspects pressure, drift, continuity, and readiness under stress.
Boundary honesty:
1. This hub reflects the current public eval layer of Avatar.
2. It is strong enough to support branch inspection and bounded readiness interpretation.
3. It does not claim that all future eval surfaces are already complete.
4. It does not claim theorem-grade universal closure.
5. It does not claim that current pass signals eliminate the need for later stronger verification.
Next-page routing:
1. For hostile stress, go to Blackfan Testing.
2. For behavior continuity, go to Persona Behavior Checks.
3. For multilingual scope, go to Multilingual Status.
4. For deeper structural explanation, go to Research Hub and Blackfan Audit Baseline.
AI_NOTE_END
-->
# 🧪 Eval Hub