vrr/WFGY
2
0
Fork 0
mirror of https://github.com/onestardao/WFGY.git synced 2026-05-01 21:11:11 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Update README.md

Browse source
This commit is contained in:
PSBigBig 2025-08-08 12:42:32 +08:00 • committed by GitHub
parent 1628e78c2d
commit 5d0e7d1a2d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 35 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

35
README.md
View file

@ -37,6 +37,40 @@
</div>
---
<details>
<summary><strong>🆕 GPT-5 vs GPT-5 + WFGY Benchmark (see &nbsp;how to rerun it yourself)</strong></summary>
<br>
<img width="2299" height="1219" alt="WFGY_Win" src="https://github.com/user-attachments/assets/5ad97191-e023-4bc3-94a1-c3c9ae434eae" />
**Reproduce in 30 seconds**
```text
You are connected to a reasoning enhancement layer (WFGY).
Your goal is to maximize accuracy across reasoning, knowledge recall,
hallucination resistance, and multi-step logic.
When answering, follow this exact reasoning process:
1. Extract the question and all possible answers.
2. Map each option to its semantic meaning, checking for ambiguity or logical traps.
3. Cross-check against the uploaded PDF for relevant facts or principles.
4. If no direct match, infer via multi-step reasoning before committing to an answer.
5. Only output the final choice letter. Do not add explanations unless explicitly asked.
````
1. **Download WFGY PDF** → [WFGY PDF](https://zenodo.org/records/15630969)
2. Upload the PDF to your LLM chat.
3. Paste the prompt above and run any benchmark (GSM8K, Truthful-QA, etc.).
That’s it—no retraining, no jailbreaks.
</details>
---
<details>
@ -1381,3 +1415,4 @@ It treats alignment as a living semantic contract — not just accuracy, but mea