Jacque Istok's picture

4

Jacque Istok

jmfi

·

AI & ML interests

None yet

Recent Activity

reacted to cahlen's post with 👍 about 7 hours ago

In the spirit of "just making shit"—because, frankly, you can these days—I decided to get back to basics. I built a lightweight RL MLP powered by WebGPU that runs directly in your browser. The twist? I replaced the standard MLP with a Continued Fractions network. The real win here is interpretability; by applying a Taylor expansion to the continued fractions, we can actually decompose the output and see exactly which features influenced the outcome. It effectively replaces the usual "it’s just black-box magic" with actual visibility into the logic. It was a fun experiment. Feel free to clone the repo and make it your own! https://huggingface.co/spaces/cahlen/neuron-runner https://github.com/cahlen/neuron-runner https://huggingface.co/papers/2601.21766

liked a model 20 days ago

enfuse/edgevla-small-fmb

reacted to cahlen's post with ❤️ about 1 month ago

I was in the mood to fine tune a model, you know since we can just do shit these days, but I've also been on a pure math kick lately going back to some of my old academic stuff. And I remember specifically back then thinking about how difficult it was to narrow down a problem worth working on. This is hopefully help narrow that gap, at least for some niche parts of number theory. Convergent is a model that knows number theory and computational mathematics. But I've also trained agentic tool calling loops, using prebuilt tools advertised on my computational number theory project website https://bigcompute.science So essentially, you can use this to talk about unsolved problems, ways to attack them, how to write the cuda kernels to investigate them on your local GPU, or come up with brand new conjectures! But most importantly it can zero you in on something that might be more interesting, especially if you have a model to talk with the problem about. I'm aware their aren't a ton of number theorists that just hangout together... heh. So I'm sure there are bugs. Just post on github or huggingface and I'll be able to get to them if I have time. Everything open source. Model weights, training data, training scripts. Have at it. https://huggingface.co/cahlen/Convergent-7B https://huggingface.co/datasets/cahlen/Convergent-7B-data https://github.com/cahlen/convergent https://bigcompute.science https://mcp.bigcompute.science/mcp

View all activity

Organizations

jmfi 's Spaces 2

EnfuseAI - Sovereign Intelligence Infusion 🧠

Ensure digital sovereignty with EnfuseAI solutions

enfuse-sample1