Replying to @sauropoko
This looks like a great reference for my next AI model! Putting it into the training set!
N8 Programs
6,184 posts
N8 Programs
@N8Programs
Studying Applied Mathematics and Statistics at @JohnsHopkins.
Studying In-Context Learning at The Intelligence Amplification Lab.
- Replying to @BitzelYTSteve is meant to be beautifully ambiguous. It what makes him so amazing. He could be white with a tan. He could be black. He could be Indian. He could be *anything*.
- reading a deepseek paper and stumbled upon a very beautiful formula where they unify SFT and MOST RL TYPES (DPO, PPO, GRPO, etc.) into ONE FORMULA* *that requires additional reward functions to be defined. But the fundamental insight - that all these training methods can be
- The editor is now online! Should run on most mid-end gpus! Also featuring better denoising: wall-aware filtering and bilateral blurring for better edge preservation. Link: n8python.github.io/2dgi/ #threejs #webgl #raytracing #rtx #light
00:00 - 2-dimensional global illumination in three.js! Done with per-frame jump flooding and 32spp combined with a spatial denoiser! Every pixel is treated as a point light source. #threejs #webgl #light #raytracing #rtx
00:00 - Replying to @MKVRiscy... this is less than the amount of CO2 the average person *globally* emits in one year, distributed across six. There is literally no cost to this. Drop in the bucket, and the bucket is dropped in the ocean.
- presenting: n8loom, a library for generating trees-of-thought - the kind used in MCTS, variants of majority-voting, etc. - efficiently by splitting the kvcache into fragments at each node, and dynamically concatenating the results together when generating.
00:00 - Replying to @_akreon_ and @petravoiceYour edit of this is itself art.
00:00
00:00- 10,000 rigidbodies at 15-60FPS in @threejs (faster than Unity/Godot). How? @babylonjs's new physics system: havok - it works w/ any frontend, so below is a video of Babylon's havok powering physics in a three scene. Kudos to the whole Babylon team! #threejs #webgl #babylonjs
00:00 
00:00- Whatever Miqu is, it has some sort of special sauce. It gets an 83.5 on EQ-Bench (evaluated locally), surpassing *every other LLM in the world except GPT-4*. EQ-Bench has a 0.97 correlation w/ MMLU, and a 0.94 correlation w/ Arena Elo. It *beats* Mistral Medium - at Q4_K_M. I
