Upload README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Chain-of-Thought as Sequential Conditional Probability Constriction
|
| 2 |
+
|
| 3 |
+
## Author
|
| 4 |
+
**Kevin T.N**
|
| 5 |
+
Email: jkdkr2439@gmail.com
|
| 6 |
+
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
## What is this?
|
| 10 |
+
|
| 11 |
+
This repository contains a **theoretical / mathematical interpretation of Chain-of-Thought (CoT) prompting** for large language models.
|
| 12 |
+
|
| 13 |
+
Instead of treating CoT as:
|
| 14 |
+
- "reasoning"
|
| 15 |
+
- "thinking step by step"
|
| 16 |
+
- or a cognitive mechanism
|
| 17 |
+
|
| 18 |
+
this work formalizes CoT as:
|
| 19 |
+
|
| 20 |
+
> **A sequential conditional probability constriction process**,
|
| 21 |
+
> where each intermediate token re-conditions the model distribution,
|
| 22 |
+
> progressively narrowing entropy and steering probability mass.
|
| 23 |
+
|
| 24 |
+
In short:
|
| 25 |
+
- CoT does **not** add intelligence
|
| 26 |
+
- It reshapes **how probability is conditioned and sampled**
|
| 27 |
+
|
| 28 |
+
---
|
| 29 |
+
|
| 30 |
+
## Who is this for?
|
| 31 |
+
|
| 32 |
+
This work may be useful if you are:
|
| 33 |
+
- A researcher working on **LLM interpretability**
|
| 34 |
+
- Interested in **reasoning vs. probability steering**
|
| 35 |
+
- Designing **prompting, inference, or sampling strategies**
|
| 36 |
+
- Thinking about CoT, ToT, ReAct from a **theoretical / information-theoretic** angle
|
| 37 |
+
|
| 38 |
+
This is **not**:
|
| 39 |
+
- a benchmark paper
|
| 40 |
+
- a performance hack
|
| 41 |
+
- a production recipe
|
| 42 |
+
|
| 43 |
+
It is a **conceptual tool**.
|
| 44 |
+
|
| 45 |
+
---
|
| 46 |
+
|
| 47 |
+
## How to use this work
|
| 48 |
+
|
| 49 |
+
You are encouraged to:
|
| 50 |
+
- Reuse the mathematical framing
|
| 51 |
+
- Adapt the notation to your own context
|
| 52 |
+
- Extend it with:
|
| 53 |
+
- Bayesian inference
|
| 54 |
+
- filtering / particle methods
|
| 55 |
+
- causal or SCM analysis
|
| 56 |
+
- empirical validation
|
| 57 |
+
|
| 58 |
+
If you find it useful:
|
| 59 |
+
- **Please credit or shout out the author** (Kevin T.N)
|
| 60 |
+
- You do **not** need permission
|
| 61 |
+
- You **should** adapt and update it to your own framing
|
| 62 |
+
|
| 63 |
+
This work is intentionally open and unfinished.
|
| 64 |
+
|
| 65 |
+
---
|
| 66 |
+
|
| 67 |
+
## Citation / Attribution (informal)
|
| 68 |
+
|
| 69 |
+
If you reference this idea in writing or discussion, a simple attribution is enough:
|
| 70 |
+
|
| 71 |
+
> “Based on a probabilistic interpretation of Chain-of-Thought as sequential conditional distribution constriction by Kevin T.N.”
|
| 72 |
+
|
| 73 |
+
Formal citation is optional.
|
| 74 |
+
|
| 75 |
+
---
|
| 76 |
+
|
| 77 |
+
## License
|
| 78 |
+
|
| 79 |
+
This project is released under the **MIT License**.
|
| 80 |
+
|
| 81 |
+
You are free to:
|
| 82 |
+
- use
|
| 83 |
+
- copy
|
| 84 |
+
- modify
|
| 85 |
+
- merge
|
| 86 |
+
- publish
|
| 87 |
+
- distribute
|
| 88 |
+
- sublicense
|
| 89 |
+
|
| 90 |
+
with **no warranty**, as long as the license notice is included.
|
| 91 |
+
|
| 92 |
+
---
|
| 93 |
+
|
| 94 |
+
## MIT License
|
| 95 |
+
|
| 96 |
+
MIT License
|
| 97 |
+
|
| 98 |
+
Copyright (c) 2026 Kevin T.N
|
| 99 |
+
|
| 100 |
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
| 101 |
+
of this software and associated documentation files (the "Software"), to deal
|
| 102 |
+
in the Software without restriction, including without limitation the rights
|
| 103 |
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
| 104 |
+
copies of the Software, and to permit persons to whom the Software is
|
| 105 |
+
furnished to do so, subject to the following conditions:
|
| 106 |
+
|
| 107 |
+
The above copyright notice and this permission notice shall be included in all
|
| 108 |
+
copies or substantial portions of the Software.
|
| 109 |
+
|
| 110 |
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
| 111 |
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
| 112 |
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
| 113 |
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 114 |
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 115 |
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 116 |
+
SOFTWARE.
|