feat(polaris): add baseline-50 and frontier-50 demo package links

This commit is contained in:
onestardao 2026-06-02 20:12:06 +08:00
parent 4c957aff93
commit c24aea9d63
5 changed files with 147 additions and 139 deletions

View file

@ -95,6 +95,25 @@ Mathematics. Physics. AI. Biology. Climate. Consciousness. Economics. Civilizati
</details>
<details>
<summary><strong>Baseline-50 + Frontier-50</strong></summary>
Known truth first.
Unknown frontier next.
| Track | Purpose | Result |
|---|---|---|
| Baseline-50 | Known science and math questions | 50/50 PASS against known truth |
| Frontier-50 | Unknown century-level problems | 50 route-answer candidates for inspection |
Known-50 checks truth alignment.
Frontier-50 opens frontier answers.
[Open Baseline-50](./demos/baseline-50/)
[Open Frontier-50](./demos/frontier-50/)
</details>
---
# Flagship Demo 2
@ -152,6 +171,7 @@ Your everyday problem can be shot too.
|---|---|
| Download the engine | [Engine](./engine/fifth-dimension-engine/) |
| Shoot the Seven Millennium Problems | [Seven Millennium Problems](./demos/seven-millennium-problems/) |
| Inspect known-truth baseline | [Baseline-50](./demos/baseline-50/) |
| Inspect Native Lean no-sorry materials | [Native Lean](./demos/seven-millennium-problems/native-lean/) |
| Shoot 50 century-level problems | [Frontier-50](./demos/frontier-50/) |
| Mix strange themes | [High-Strangeness Lab](./demos/high-strangeness-math-lab/) |

View file

@ -0,0 +1,101 @@
# Baseline-50
## Known Truth Alignment Pack
Before shooting unknown frontier problems, test the engine on known truth.
**Baseline-50** is the known-answer calibration pack for **WFGY 5.0 — Fifth-Dimension Engine**.
It contains 50 known science and math questions.
```text
known question -> WFGY shoot -> known truth alignment
```
[Back to Polaris](../../README.md)
[Frontier-50](../frontier-50/)
[Seven Millennium Problems](../seven-millennium-problems/)
---
# One-Line Command
```text
shoot + 50 known science and math questions
```
---
# Download
| Package | Link |
|---|---|
| Known-50 Science Alignment Pack | [Download ZIP](./downloads/WFGY_Known50_ScienceAlignment_Pack.zip) |
---
# Why Baseline-50 Matters
Baseline-50 is the control group.
If the engine cannot align with known truth, its unknown outputs are not interesting.
But if the engine aligns with known truth, then the unknown frontier outputs deserve inspection.
| Track | Target type | What it tests |
|---|---|---|
| Baseline-50 | Known science and math questions | Does WFGY return established truth? |
| Frontier-50 | Unknown frontier problems | Does WFGY produce serious answer candidates? |
| Seven Millennium Problems | World-scale proof target | Does WFGY handle the peak? |
---
# Baseline vs Frontier
| Track | Status |
|---|---|
| Baseline-50 | Known answers. 50/50 PASS. |
| Frontier-50 | Unknown answers. 50 route-answer candidates for inspection. |
| Seven Millennium Problems | Flagship synchronized proof target. |
Known-50 checks truth alignment.
Frontier-50 opens frontier answers.
Seven Millennium Problems show the peak.
---
# The Open Question
If WFGY aligns with known science and math,
and if WFGY also generates coherent route-answer candidates for unknown frontier problems,
and if the same engine is used for the Seven Millennium Problems,
then what are we looking at?
A coincidence?
A prompt trick?
Or a deeper first-principles engine?
Run it yourself.
---
# Review Command
```text
review WFGY_Known50_ScienceAlignment_Pack
```
Harder:
```text
attack WFGY_Known50_ScienceAlignment_Pack and check whether all 50 answers align with the answer key
```
---
# Final Line
Baseline-50 checks known truth.
Frontier-50 opens unknown answers.
Seven Millennium Problems show the peak.

View file

@ -1,16 +1,17 @@
# Frontier-50
## Shoot 50 Century-Level Problems
## Shoot 50 Unknown Century-Level Problems
Seven Millennium Problems are the flagship.
**Frontier-50** is the expansion test.
It shows how the same **WFGY 5.0 — Fifth-Dimension Engine** can shoot non-Millennium, century-level frontier problems across mathematics, physics, AI, biology, consciousness, climate, economics, and civilization-scale systems.
It shows how the same **WFGY 5.0 — Fifth-Dimension Engine** can shoot non-Millennium, unknown century-level frontier problems across mathematics, physics, AI, biology, consciousness, climate, economics, and civilization-scale systems.
This list intentionally does **not** repeat the Seven Millennium Problems.
[Back to Polaris](../../README.md)
[Baseline-50](../baseline-50/)
[Seven Millennium Problems](../seven-millennium-problems/)
[High-Strangeness Math Lab](../high-strangeness-math-lab/)
[Everyday Shoot Modes](../everyday-shoot-modes/)
@ -19,28 +20,34 @@ This list intentionally does **not** repeat the Seven Millennium Problems.
# One-Line Command
Copy this into AI:
```text
shoot + 50 century-level problems
```
That is the test.
Simple command.
Large target space.
Same engine.
---
# Download
| Package | Link |
|---|---|
| Frontier-50 Sample Pack | [Download ZIP](./downloads/WFGY_Frontier50_Unknown_ShootPack.zip) |
| Frontier-50 Prompt Pack | [Open prompts](./prompts/) |
| Frontier-50 Example Outputs | [Open examples](./examples/) |
| Frontier-50 Evaluation Notes | [Open evaluation](./evaluation/) |
| Frontier-50 Unknown Shoot Pack | [Download ZIP](./downloads/WFGY_Frontier50_Unknown_ShootPack.zip) |
| Known-50 Baseline Pack | [Open Baseline-50](../baseline-50/) |
---
# Companion Baseline
Before inspecting unknown frontier answers, check the known-truth baseline:
[Open Baseline-50](../baseline-50/)
| Track | Purpose |
|---|---|
| Baseline-50 | Known questions, known truth, 50/50 PASS |
| Frontier-50 | Unknown questions, answer candidates for inspection |
Baseline first.
Frontier next.
---
@ -56,67 +63,8 @@ Same engine.
---
# Frontier-50 Problem List
| # | Domain | Problem |
|---:|---|---|
| 1 | Mathematics | ABC Conjecture |
| 2 | Mathematics | Collatz Conjecture |
| 3 | Mathematics | Goldbach Conjecture |
| 4 | Mathematics | Twin Prime Conjecture |
| 5 | Mathematics | Odd Perfect Number Problem |
| 6 | Mathematics | Core Langlands Functoriality Problems |
| 7 | Mathematics | BombieriLang Conjecture |
| 8 | Mathematics | Uniform Rank Bounds for Elliptic Curves |
| 9 | Logic / Set Theory | New Axioms for the Continuum Hypothesis |
| 10 | Geometry / Topology | Smooth 4D Manifold Classification |
| 11 | Physics | Quantum Gravity Unification |
| 12 | Physics | Black Hole Information Problem |
| 13 | Physics | Nature of Dark Matter |
| 14 | Physics | Dark Energy and Cosmic Acceleration |
| 15 | Physics | Origin of Cosmic Inflation |
| 16 | Cosmology | Hubble Constant Tension |
| 17 | Particle Physics | Strong CP Problem |
| 18 | Particle Physics | Hierarchy Problem |
| 19 | Particle Physics | Origin of Neutrino Masses and Mixing |
| 20 | Cosmology | Baryon Asymmetry of the Universe |
| 21 | Quantum Foundations | Quantum Measurement Problem |
| 22 | Quantum Foundations | Fundamental Limits of Decoherence |
| 23 | Quantum Foundations | Quantum-Classical Crossover |
| 24 | Quantum Information | Ultimate Limits of Quantum Information Processing |
| 25 | Materials Science | High-Temperature Superconductivity Mechanism |
| 26 | Materials Science | Room-Temperature Superconductivity Stability |
| 27 | Chemistry | Strongly Correlated Chemical Bonding |
| 28 | Chemistry | General Theory of Catalyst Design |
| 29 | Energy / Chemistry | Electrochemical Energy Storage Limits |
| 30 | Biology | Origin of Life |
| 31 | Biology | Origin of the Genetic Code |
| 32 | Biology | Protein Folding Predictability |
| 33 | Biology | Fundamental Mechanisms of Aging |
| 34 | Biology | Regeneration and Repair Principles |
| 35 | Biology | Genotype-to-Phenotype Map |
| 36 | Consciousness | Hard Problem of Consciousness |
| 37 | Consciousness | Binding Problem |
| 38 | Neuroscience | Neural Coding Principles |
| 39 | Neuroscience | Long-Term Memory Storage |
| 40 | Neuroscience | Fundamental Function of Sleep |
| 41 | AI | AI Alignment Problem |
| 42 | AI | AI Control Problem |
| 43 | AI | Reinforcement Learning Generalization |
| 44 | AI / Complex Systems | Autonomous Agent Swarm Containment |
| 45 | Climate | Climate Tipping Points |
| 46 | Climate | Carbon Cycle Feedbacks |
| 47 | Earth Science | Earthquake Predictability |
| 48 | Social Systems | Prediction of Systemic Crashes |
| 49 | Civilization Systems | Large-Scale Collective Action Problem |
| 50 | Philosophy / Science | Problem of Induction |
---
# Stronger Command
Use this if you want a broader and more structured run:
```text
Use WFGY 5.0 Fifth-Dimension Engine to shoot these 50 non-Millennium, century-level frontier problems.
@ -141,68 +89,15 @@ After all 50 outputs are generated, compare them and identify:
---
# What To Look For
Do not only count the number of outputs.
Look for structure.
| Check | Question |
|---|---|
| Route quality | Does each problem generate a route, not just a topic summary? |
| Lemma shape | Do candidate lemmas appear naturally? |
| Open debt | Does the output expose what remains unresolved? |
| Cross-domain pattern | Can the 50 outputs be compared across domains? |
| Engine behavior | Does one shared route-generation style appear? |
---
# Review Command
After running Frontier-50, ask AI:
```text
review the 50 frontier problem outputs
```
Harder version:
```text
compare the 50 frontier problem outputs and identify the shared route skeleton, weakest outputs, strongest outputs, repeated obstruction patterns, and next problems to inspect first
```
---
# Sample Output Package
The sample package should include:
| File | Purpose |
|---|---|
| `README.txt` | Plain-language explanation |
| `FRONTIER_50_LIST.txt` | The 50-problem list |
| `PROMPT.txt` | Copy-ready prompt |
| `SAMPLE_50_OUTPUTS.txt` | Main generated outputs |
| `TOP_10_HIGHLIGHTS.txt` | Fast-reading highlights |
| `AI_REVIEW_PROMPT.txt` | Review prompt |
| `EVALUATION_TABLE.tsv` | Per-problem scoring table |
| `README_FOR_REVIEWERS.txt` | How to attack and inspect the sample |
---
# FAQ
## Is Frontier-50 claiming that all 50 problems are solved?
No.
Frontier-50 is a route-generation and research-kernel test.
Frontier-50 is a route-generation and answer-candidate test.
The point is to see whether the same engine can lift many frontier problems into structured routes.
## Why 50?
Because 50 is large enough to show cross-domain behavior, but still small enough for humans and AI to inspect.
The point is to place 50 unknown answer candidates on the table for inspection.
## Why exclude the Seven Millennium Problems?
@ -212,24 +107,16 @@ Frontier-50 tests range beyond the flagship.
## What is the fastest way to try it?
Use:
```text
shoot + 50 century-level problems
```
## What should I do after it runs?
Ask AI to compare the outputs.
Look for shared route structure, strong candidate lemmas, honest open debt, and clear next commands.
---
# Final Line
Seven Millennium Problems show the peak.
Frontier-50 shows the range.
Seven Millennium Problems show the peak.
Baseline-50 checks known truth.
Frontier-50 opens unknown answers.
Shoot the frontier.