anima-core

6 ms·

1.
by anima-core 7mo ago
https://www.linkedin.com/posts/ryanshamim_deeptechpioneers-h...
2.
All of you doubted me, and said I was hallucinating on AI(ryanshamim.substack.com)
1 points by anima-core 7mo ago|1 comments
3.
We keep making transformers faster. What if we don't need them at all?(zenodo.org)
1 points by anima-core 9mo ago|1 comments
4.
by anima-core 9mo ago
We keep spending enormous effort making transformers run faster. Quantization. Pruning. Speculative decoding. Better kernels. Better hardware. All of that assumes the same thing: that every request should run the model. I’ve been working on
5.
A One-Shot Catastrophe Avoidance Benchmark for RL Agents(zenodo.org)
2 points by anima-core 9mo ago|1 comments
6.
by anima-core 9mo ago
This paper introduces a minimal benchmark for testing whether an RL agent can learn a permanent safety constraint from a single catastrophic event. The protocol uses standard MiniGrid LavaCrossing environments, fixed seeds, and forbids any
7.
Semantic Field Execution: Decoupling Transformers from Runtime Inference(zenodo.org)
1 points by anima-core 9mo ago|1 comments
8.
by anima-core 9mo ago
This is a short position paper that asks a narrow systems question: what changes if large transformers are removed from the runtime inference loop entirely? The paper introduces Semantic Field Execution (SFE), an inference substrate in whic
9.
The Inference Efficiency Paradox(substack.com)
2 points by anima-core 9mo ago|0 comments
10.
by anima-core 9mo ago
Haha, fair play, sir. If anything induces altered states around here, it’s probably late-night debugging rather than substances. I'm here to talk experiments, code, and results. Im ready to dive into that whenever you guys are.
11.
by anima-core 9mo ago
I appreciate this take. I largely agree with the framing, and I think this is closer to the intended reading than some of the more heated responses in the thread. (I'm understanding this is whats expected in the forum, and now I welcom
12.
by anima-core 9mo ago
I can see you've put real thought into your critique, and while I definitely disagree with several conclusions, I appreciate the seriousness of the discussion. Hopefully this is a good faith discussion, and we can keep it that way. Let
13.
by anima-core 9mo ago
I appreciate this framing a lot. It is actually close to how I think about the result internally. The paper focuses on the geometric behavior of intermediate representations, and classification is the cleanest setting to study that. Generat
14.
by anima-core 9mo ago
What are you a psychiatrist?
15.
by anima-core 9mo ago
Thank you for the thoughtful comments. Really. This is actually the most constructive feedback in the thread so far. A few clarifications. 1. On the LaTeX citations and figure references That part is definitely on me. I never used LaTeX bef
16.
by anima-core 9mo ago
A few clarifications, since most of the points here come from asking LLMs to summarize the repo rather than running the code directly. 1. The teacher only runs during field extraction. That step is offline. Once the fields are saved, the tr
17.
by anima-core 9mo ago
That's not how the method works... The full transformer is only needed once to extract the activation fields. That step can even be done offline. Then the teacher can be discarded entirely. The compression result refers to the size of
18.
by anima-core 9mo ago
When someone shifts from engaging with the actual results to attacking the person, it usually tells you more about their internal state than about the work itself. I'm glad I have a new fan though.
19.
by anima-core 9mo ago
Oh, so you didnt run the repo and remembered something that you read once that looked like it matched. This contribution is meaningless. The simplest way to resolve any doubt is to run the code. Every result in the paper comes from reproduc
20.
by anima-core 9mo ago
The substack isnt what was supposed to be evaluated, it was the repo. That's creative writing and the repo is sciencetific. Two different things. One has nothing to do with the other. The technical direction here is straightforward, al
21.
by anima-core 9mo ago
That limitation is already accounted for in how the title is meant to be read. The 224× compression result is specifically about the structure of intermediate activations on classification tasks. The paper makes that explicit in multiple pl
22.
by anima-core 9mo ago
I guess my "vibe" is just better than your coding :)... Let me explain a few things, if you will. A few clarifications so the discussion stays aligned with what the experiment is actually measuring. 1. The HellaSwag “binary collap
23.
by anima-core 9mo ago
As a follow up just to refresh your memory: “Attention Is All You Need” (Vaswani et al., 2017) Length: 11 pages of main content, 5 pages of references and appendix 2. The first GPT paper (Radford et al., 2018) Length: 12 pages 3. BERT (Devl
24.
by anima-core 9mo ago
The memory story is actually much simpler than it looks. The teacher still has to be loaded at training time, so the footprint is whatever the original model uses. Again, the compression doesn't shrink the teacher. It produces a small
25.
by anima-core 9mo ago
No, the compression result doesn't mean the original 64 GB model can run on a 292 MB card. The teacher model isn’t the thing thats compressed. It still needs to be loaded during training. What gets small is the student. The tiny head t
26.
by anima-core 9mo ago
I appreciate you taking the time to resond, brother. Let me clarify a few things because your interpretation misses the actual structure of the work. The paper is short on purpose. It's not meant as a full architecture release. It&#x27
27.
by anima-core 9mo ago
I’m actually building the system-level approach this memo hints at. I’m not from a lab or an academic group, but I’ve been working on a post-transformer inference method where you extract a low-rank “meaning field” from a frozen Llama-70B l
28.
Post-transformer inference: 224× compression of Llama-70B with improved accuracy(zenodo.org)
72 points by anima-core 9mo ago|56 comments
29.
by anima-core 9mo ago
I’ve been working independently on a method that replaces full-transformer inference with a low-rank “meaning field” extracted from internal activations. The core result: a frozen Llama-3.3-70B can be distilled into a 256-dimensional field