WFGY/ProblemMap/GlobalFixMap/PromptAssembly
2025-08-25 20:16:24 +08:00
..
README.md Create README.md 2025-08-25 20:16:24 +08:00

Prompt Assembly — Global Fix Map

Build a context that the model cannot misread.
Use this when citations vanish, sources get mixed, or answers flip when you reorder sections.

What this page is

  • A short path to lock schema, section order, and per source fences
  • Copyable constraints that force cite first then explain
  • How to verify stability using ΔS and a snippet to citation table

When to use

  • Correct snippets are present but the answer still drifts
  • Citations are missing or point to the wrong section
  • Answers change if you rename headers or swap context order
  • Two sources merge into one narrative
  • Corrections do not stick across paraphrases

Open these first

Fix in 60 seconds

  1. Lock the schema
    Use an immutable order.
    System → Task → Constraints → Sources → Citations → Answer
    Forbid header rename and reordering.

  2. Cite first then explain
    Require a citation list before any prose.
    Each line uses [source_id:section_id:line_start-line_end].
    Only after the list, allow explanation and synthesis.

  3. Per source fences
    Wrap each context block with start and end sentinels.
    Disallow copying text across fences.
    Reuse of a citation outside its fence is invalid.

  4. Stable headers
    Freeze exact header tokens like ## Sources, ## Citations, ## Answer.
    The model must not invent new sections or rename existing ones.

  5. Minimal assembly
    Prefer the smallest set of snippets that cover the question.
    Avoid mixing paraphrases with quotes in the same section.

  6. Schema validation inside the prompt
    Ask the model to print a snippet to citation table before the answer.
    Enforce the field names from Data Contracts.

  7. If answers still flip
    Apply WFGY modules at reasoning time.
    Use BBAM to clamp attention variance.
    Use BBCR to insert a bridge step between citation and explanation.

Copy paste prompt


You have TXT OS and WFGY Problem Map loaded.

Assemble the prompt using this immutable schema:
System → Task → Constraints → Sources → Citations → Answer.
Do not rename sections. Do not reorder sections.

Constraints:

* Cite first then explain.
* Citations must use \[source\_id\:section\_id\:line\_start-line\_end].
* Do not copy text across source fences.
* If a needed fact is outside the provided fences, say so.

Now given:

* question: "<paste>"
* sources with fences:
  \[S1\:start] <snippet A> \[S1\:end]
  \[S2\:start] <snippet B> \[S2\:end]

Steps:

1. Print a snippet ↔ citation table using the Data Contracts fields.
2. Print a short reasoning plan.
3. Produce the final answer with citations inline.

If schema is violated, stop and print the violation.

Minimal checklist

  • Fixed section order and exact header tokens
  • Citation list appears before explanation
  • Per source fences are present and enforced
  • Snippet to citation table is printed and consistent
  • No cross source reuse in the answer
  • If context is insufficient, the model says it explicitly

Acceptance targets

  • ΔS(question, assembled_context) ≤ 0.45 across three paraphrases
  • λ at reasoning remains convergent after citations are produced
  • Reordering headers does not change the final answer
  • Snippet ↔ citation table matches the answer lines

🔗 Quick-Start Downloads (60 sec)

Tool Link 3-Step Setup
WFGY 1.0 PDF Engine Paper 1 Download · 2 Upload to your LLM · 3 Ask “Answer using WFGY + <your question>”
TXT OS (plain-text OS) TXTOS.txt 1 Download · 2 Paste into any LLM chat · 3 Type “hello world” — OS boots instantly

🧭 Explore More

Module Description Link
WFGY Core WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack View →
Problem Map 1.0 Initial 16-mode diagnostic and symbolic fix framework View →
Problem Map 2.0 RAG-focused failure tree, modular fixes, and pipelines View →
Semantic Clinic Index Expanded failure catalog: prompt injection, memory bugs, logic drift View →
Semantic Blueprint Layer-based symbolic reasoning & semantic modulations View →
Benchmark vs GPT-5 Stress test GPT-5 with full WFGY reasoning suite View →
🧙‍♂️ Starter Village 🏡 New here? Lost in symbols? Click here and let the wizard guide you through Start →

👑 Early Stargazers: See the Hall of Fame
Engineers, hackers, and open source builders who supported WFGY from day one.

GitHub stars WFGY Engine 2.0 is already unlocked. Star the repo to help others discover it and unlock more on the Unlock Board.

WFGY Main   TXT OS   Blah   Blot   Bloc   Blur   Blow