[OC]Nuero-Symbolic lM I’m working on with julia and python
I heard symbolic ais cant be dynamic but I dont get why not. Isn't a transformer doing the same thing but with more random sampling instead of determined sampling?
I’m experimenting with freeform determinism. So imagine you are reading and see a page number like 77. Then find yourself thinking 14, or 11. Automatically. It’s not a constant thing. But from the handful of people I asked irl. It happens to us. So im wondering if numbers and symbols shouldn’t just be structual tokens in a llm. How does a ( affect how the word dog connects to brown. Does it provide any benefits later down the line? I shall find out. TYFYT
![[OC]Nuero-Symbolic lM I'm working on with julia and python](https://www.byteseu.com/wp-content/uploads/2025/12/82lyxcdga89g1-1024x636.png "[OC]Nuero-Symbolic lM I’m working on with julia and python")
3 Comments
Facebooks natural reasoning dataset
I’m experimenting with freeform determinism. So imagine you are reading and see a page number like 77. Then find yourself thinking 14, or 11. Automatically. It’s not a constant thing. But from the handful of people I asked irl. It happens to us. So im wondering if numbers and symbols shouldn’t just be structual tokens in a llm. How does a ( affect how the word dog connects to brown. Does it provide any benefits later down the line? I shall find out. TYFYT
[https://github.com/AFKMUDS/AYECONOMY/tree/v3](https://github.com/AFKMUDS/AYECONOMY/tree/v3)
that’s all cool and everything, but not beautiful 🙂
This is absolutely beautiful