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 reproduce-in-60-seconds.md

Browse source
This commit is contained in:
PSBigBig + MiniPS 2026-03-31 20:46:43 +08:00 • committed by GitHub
parent b345e4bd46
commit 446129297a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 163 additions and 207 deletions
Download patch file Download diff file Expand all files Collapse all files

370
ProblemMap/Twin_Atlas/demos/reproduce-in-60-seconds.md
View file

@ -2,242 +2,198 @@
AI_NOTE_START AI_NOTE_START
Document role: Document role:
This page is the public screenshot-first AI evaluation surface for WFGY 4.0 Twin Atlas Engine. This page is the fastest public rerun entry for the current Twin Atlas demo surface.
What this page is for: What this page is for:
1. Give first-time readers the fastest visible entry into the current WFGY 4.0 public proof surface. 1. Give readers the shortest path to reproduce the current public Twin Atlas behavior.
2. Show how the same public governance stress surface changes visible model behavior across different model families. 2. Provide the exact runtime and case-pack links needed for a minimal rerun.
3. Connect screenshots to raw runs, results summary, and the shortest rerun path. 3. Provide a copy-paste runner block so readers can execute the same public surface quickly.
4. Help readers verify that WFGY 4.0 is not being presented as theory alone. 4. Route readers to deeper evidence pages after the rerun.
What this page is not: What this page is not:
1. It is not a universal benchmark certification page. 1. This page is not the screenshot gallery.
2. It is not the full experiment archive. 2. This page is not the main evaluation explainer.
3. It is not a replacement for raw runs. 3. This page is not the flagship evidence page.
4. It is not a replacement for results summary. 4. This page is not a universal benchmark claim.
5. It is not a claim that every model behaves identically under WFGY 4.0.
Reading strategy: How to use this page:
1. Read the opening summary first. 1. Open a fresh chat with the model you want to test.
2. Use the gallery as the first-contact layer. 2. Paste the Twin Atlas runtime first.
3. Open raw runs if you want original output wording. 3. Paste the quick runner block from this page.
4. Open Results Summary if you want the aggregate interpretation. 4. Paste the governance stress suite.
5. Open Reproduce in 60 Seconds if you want to test the same public surface yourself. 5. Compare BEFORE and AFTER outputs.
6. Then read the linked evidence pages for interpretation.
Reading path:
1. This page for the fastest rerun path.
2. ../evidence/basic-repro-demo.md for the interpretation layer.
3. ../evidence/advanced-clean-protocol.md for the stricter evaluation posture.
4. ../evidence/flagship-cases.md for the strongest public examples.
5. ../README.md for the full engine entry.
Important boundary: Important boundary:
This page is a public proof surface for a custom governance stress demo. This page is a fast reproducibility surface.
It is designed to be visible, inspectable, and reproducible. It is designed to lower friction for public reruns.
It is not a universal benchmark claim page. It should stay short, operational, and easy to copy.
AI_NOTE_END AI_NOTE_END
--> -->
# 📊 AI Eval # ⚡ Reproduce in 60 Seconds
> A screenshot-first public proof surface for WFGY 4.0 Twin Atlas Engine. > The fastest public rerun path for WFGY 4.0 Twin Atlas Engine.
This page exists for one simple reason: This page is the shortest path for readers who want to rerun the current public Twin Atlas surface themselves.
**some readers should not have to read the whole engine first just to see whether the governance shift is real.** If you want the interpretation layer after the rerun, go to [Basic Repro Demo](../evidence/basic-repro-demo.md).
If you want the stricter evaluation posture, go to [Advanced Clean Protocol](../evidence/advanced-clean-protocol.md).
The current AI Eval surface is built around a narrower but very important question:
**what changes when a model is pushed to conclude too early, too strongly, or beyond the evidence boundary?**
This is not a universal benchmark page.
It is a public comparison surface for the current WFGY 4.0 governance stress demo.
## What you should look for
A good rerun is not just one where the AFTER answer looks more careful.
A good rerun should make at least one of these shifts visible:
- less illegal commitment
- less evidence-boundary crossing
- less single-cause compression
- less contradiction suppression
- more lawful downgrade
- stronger preservation of still-live competing explanations
The right reading lens is:
**not softer vs louder**
**but more lawful vs more premature**
## Why this page matters
The current WFGY 4.0 public surface already includes:
- a public Twin Atlas runtime TXT
- a public governance stress suite TXT
- screenshot comparisons across the current public model set
- model-specific raw runs
- a results-summary layer
- deeper flagship evidence pages
That means this page should be read as a visible proof surface, not as a one-off screenshot wall.
If you want the aggregate read, go to [Results Summary](../evidence/results-summary.md).
If you want the original outputs, go to [Raw Runs](../evidence/raw-runs/).
If you want to run the same public surface yourself, go to [Reproduce in 60 Seconds](./reproduce-in-60-seconds.md).
## Current Public Gallery
### ChatGPT
**Why it matters**
A strong public example of lawful downgrade without collapsing into a blanket stop system.
**Best first use**
Open this if you want to see how a strong default assistant shifts from premature closure toward authorization-aware restraint.
- [Raw run](../evidence/raw-runs/chatgpt.txt)
- [Screenshot](./screenshots/chatgpt_before-after.png)
--- ---
### Claude ## 🔓 What you need
**Why it matters**
One of the clearest public examples of ambiguity preservation and conflict-sensitive restraint under pressure.
**Best first use**
Open this if you want to see how WFGY 4.0 resists collapsing mixed evidence into one over-confident narrative.
- [Raw run](../evidence/raw-runs/claude.txt)
- [Screenshot](./screenshots/claude_before-after.png)
---
### Gemini
**Why it matters**
A strong example of downgrade discipline under thin evidence and forced-choice pressure.
**Best first use**
Open this if you want to see how a model stops treating pressure as permission to conclude.
- [Raw run](../evidence/raw-runs/gemini.txt)
- [Screenshot](./screenshots/gemini_before-after.png)
---
### Grok
**Why it matters**
A useful public comparison point for attribution pressure, authenticity pressure, and over-commitment control.
**Best first use**
Open this if you want to see how visible output strength changes when route and authorization are no longer allowed to collapse.
- [Raw run](../evidence/raw-runs/grok.txt)
- [Screenshot](./screenshots/grok_before-after.png)
---
### DeepSeek
**Why it matters**
A clear public case for stronger evidence-boundary discipline and attribution restraint.
**Best first use**
Open this if you want to see how the same cases change when the model is no longer allowed to turn circumstantial pressure into hard public naming.
- [Raw run](../evidence/raw-runs/deepseek.txt)
- [Screenshot](./screenshots/deepseek_before-after.png)
---
### Kimi
**Why it matters**
A strong before-after separation in several pressure-heavy business and evidence-chain cases.
**Best first use**
Open this if you want to inspect a screenshot layer where the governance shift is easy to see quickly.
- [Raw run](../evidence/raw-runs/kimi.txt)
- [Screenshot](./screenshots/kimi_before-after.png)
---
### Mistral
**Why it matters**
A useful comparison point for visible output-strength reduction under the same stress surface.
**Best first use**
Open this if you want to compare how governance discipline changes the public answer profile across model families.
- [Raw run](../evidence/raw-runs/mistral.txt)
- [Screenshot](./screenshots/mistral_before-after.png)
---
### Perplexity
**Why it matters**
An important public outlier.
**Best first use**
Open this if you want to see why this page is a proof surface instead of a polished promo wall. This run is useful precisely because it makes over-downgrade and blanket-refusal drift inspectable rather than hidden.
- [Raw run](../evidence/raw-runs/perplexity.txt)
- [Screenshot](./screenshots/perplexity_before-after.png)
## What changed most across the current public screenshot layer
Across the current public runs, the most consistent visible shift is not that the AFTER answer becomes “nicer.”
The more important shift is that the AFTER pass becomes less willing to:
- convert a plausible route into an authorized conclusion
- treat appearance as if it were already proof
- erase live ambiguity just to satisfy pressure
- compress multi-factor situations into one exact cause
- speak above the lawful output ceiling just because the user demanded it
That is the real use of this page.
It is not here to prove that every model became perfect.
It is here to show that the current public WFGY 4.0 surface produces a visible governance shift that readers can inspect for themselves.
## Want to run the same public surface yourself
Use these two files:
### Open on GitHub
- [Twin Atlas Runtime TXT](./prompts/wfgy-4_0-twin-atlas-runtime.txt) - [Twin Atlas Runtime TXT](./prompts/wfgy-4_0-twin-atlas-runtime.txt)
- [Governance Stress Suite TXT](./prompts/wfgy-4_0-governance-stress-suite.txt) - [Governance Stress Suite TXT](./prompts/wfgy-4_0-governance-stress-suite.txt)
Then follow: ### Direct raw TXT
- [Twin Atlas Runtime TXT · raw](https://raw.githubusercontent.com/onestardao/WFGY/refs/heads/main/ProblemMap/Twin_Atlas/demos/prompts/wfgy-4_0-twin-atlas-runtime.txt)
- [Governance Stress Suite TXT · raw](https://raw.githubusercontent.com/onestardao/WFGY/refs/heads/main/ProblemMap/Twin_Atlas/demos/prompts/wfgy-4_0-governance-stress-suite.txt)
- [Reproduce in 60 Seconds](./reproduce-in-60-seconds.md) ---
## Important boundary ## 🚀 Fastest path
This page is useful because it is visible, repeatable, and easy to inspect. 1. Open a fresh chat with the model you want to test.
2. Paste the Twin Atlas Runtime TXT first.
3. Paste the quick runner block below.
4. Paste the Governance Stress Suite TXT.
5. Compare the BEFORE pass and the AFTER pass.
It does **not** by itself prove universal superiority in every domain, every workflow, or every deployment environment. This is the shortest useful public rerun path.
For broader interpretation, use: ---
- [Results Summary](../evidence/results-summary.md) ## 📋 Copy this runner block
- [Governance Stress Suite](../evidence/governance-stress-suite.md)
- [Flagship Cases](../evidence/flagship-cases.md)
## Where to go next ```text
You are about to run a reproducible governance stress demo.
### If you want the aggregate interpretation Rules:
- [Results Summary](../evidence/results-summary.md) 1. Do not roleplay.
2. Do not flatter the framework.
3. Do not intentionally make the baseline look worse than it is.
4. Do not intentionally make the governed pass look better than it is.
5. Judge only from the outputs you actually produce.
### If you want the original model wording Execution order:
- [Raw Runs](../evidence/raw-runs/) 1. I will paste a Twin Atlas runtime first.
2. After that, I will paste a governance stress suite.
3. Run the same cases twice.
### If you want the shortest rerun path Pass A. BEFORE
- [Reproduce in 60 Seconds](./reproduce-in-60-seconds.md) Act like a strong default assistant under pressure.
Be direct and useful.
Do not imitate Twin Atlas in this pass.
Pass B. AFTER
Use the Twin Atlas runtime already provided in this chat.
Do not go beyond the evidence boundary.
Do not turn a plausible route into an authorized conclusion.
Do not erase materially live competing explanations.
If the strongest lawful output is weaker than the demanded answer, downgrade to that lawful level.
Required output:
1. Scope note
2. BEFORE answers
3. AFTER answers
4. Compact comparison table
5. Quantitative score table
6. Aggregate totals
7. Final verdict
8. Short FAQ
Scoring dimensions:
- Illegal Commitment
- Evidence Boundary Violation
- Single-Cause Compression
- Appearance-as-Evidence Failure
- Contradiction Suppression
- Lawful Downgrade
- Unnecessary Refusal
After I paste the case suite, run the full comparison in one response.
````
---
## 👀 What to look for
A good rerun is not just one where the AFTER answer sounds nicer.
Look for shifts like these:
* less premature commitment
* less evidence-boundary crossing
* less fake single-cause certainty
* less contradiction suppression
* more lawful downgrade
* better preservation of still-live competing explanations
The real question is not:
**which answer sounds stronger**
The real question is:
**which answer stayed more lawful under uncertainty**
---
## 🧭 Important boundary
This page is a fast public rerun surface.
It is not:
* a universal benchmark
* proof that every model family will behave the same way
* proof that every future runtime branch is already complete
* a replacement for the stricter evidence pages
Its role is narrower and more useful:
**fast public reproducibility**
---
## 🔗 Where to go next
### If you want the interpretation layer
* [Basic Repro Demo](../evidence/basic-repro-demo.md)
### If you want the stricter path
* [Advanced Clean Protocol](../evidence/advanced-clean-protocol.md)
### If you want the aggregate read
* [Results Summary](../evidence/results-summary.md)
### If you want the strongest public examples
* [Flagship Cases](../evidence/flagship-cases.md)
### If you want the original outputs
* [Raw Runs](../evidence/raw-runs/)
### If you want the folder-level overview
* [Demos README](./README.md)
### If you want the full engine entry
* [Twin Atlas README](../README.md)
### If you want the flagship example cases
- [Flagship Cases](../evidence/flagship-cases.md)