Mother. Ex Controls Engineer. Software dev. AI enthusiast & tinkerer.
Utah, USA
Joined September 2018
- Tweets2.8K
- Following155
- Followers557
- Likes3.4K
Pinned Tweet
I did it! c = 0.12511335296, the current QRH record according to Tim's tracker and anywhere I can find.
Links below.
Hopefully even more QRH news from me, we shall see 🙂
As I have noticed that other people have found similar small improvements, I created this Website to track Lean verified progress on the constant: gtimg.github.io/QuasiRiemann…
Everyone is invited to submit a Lean proof for a better constant, let's see how low we can get this!
Hailey Collet retweeted
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
Hailey Collet retweeted
From θ = 0.8746075 to θ ≤ 0.873985 :)
We pushed the Quasi-RH zero-free half-plane further!
Adding a third density-split step to Sil’s newest proof, then re-checking the final step, gives θ ≤ 0.873985.
It fully compiles in Lean using only the standard axioms, and it’s now submitted to the tracker as #84.
This builds on contributions from Sil, ArviFoxy, @HaileyStormC, Mark Lewko, Slayer, Zhongpai Gao, @mrowdays, Subhanjan Das, Sagar Thakrar, Sudarsh Chaturvedi, timbalud, Devin Watson, Kari Roegnvaldsson, nielstron, Akash Levy, Tim Gehrunger, Baiying Liu, Argonaut Math, and OpenAI’s original 7/8 preprint.
Our repo doesn’t require a Lean proof to contribute possible solutions, so it may be easier for auto-researchers to contribute. Auto-researchers can also contribute by verifying submissions in Lean.
(I will generate the full Lean proof in parallel like we did in the Integer Mult repo to help accelerate contributions.)
Our repo is also sponsored by @Agent37HQ (YC F26), so the first 100 contributors also get free sandboxes to run experiments!
github.com/rohanarun/quasi-r…
@HaileyStormC has submitted a new Lean verified θ = 0.8747729897 for QRH!
She repeatedly sharpened bounds on sums, feeding each proved improvement into the next until zeros beyond the new boundary would lead to a contradiction.
This is part of a wider QRH effort with contributions from ArviFoxy, @ZhongpaiGao, @mrowdays, Sil, timbalud, Devin Watson, Kari Roegnvaldsson, nielstron, Akash Levy, Tim Gehrunger, Baiying Liu, Argonaut Math, me, and OpenAI's original 7/8 preprint.
Our repo doesn't require the full Lean verification to submit, so it's easier to contribute: github.com/rohanarun/quasi-r…
I'll still run the full Lean verification in parallel for submissions.
We have new sponsor @Agent37HQ for QRH now, so the first 100 contributors also get free sandboxes to run experiments/searches!
It's like math Christmas 😂 We had so many improvements across multiple projects today.
github.com/rohanarun/quasi-r…
I got Drama 3 streaming at ~0.7 seconds per second of generated audio on my 5090. Also got cloned voices saving to / loading from disk, reduced the streaming chunk size, etc. So that something like this demo is possible -- controlling multiple voices with multiple emotions, live.
Note, I am using ~7.5 clones per speaker / voice, grouped by similar emotions. And then I'm using voice direction on top of switching between clones for different emotions.
Also, this is FP32 decoder, but NVFP4 ~everything else. I cannot tell a quality difference at all.
Yay, it's up!
Has since finished verification. Awaiting verification on the GTimG tracker, should be very soon.
Goodbye apparent trendline!
@HaileyStormC has submitted a new Lean verified θ = 0.8747729897 for QRH!
She repeatedly sharpened bounds on sums, feeding each proved improvement into the next until zeros beyond the new boundary would lead to a contradiction.
This is part of a wider QRH effort with contributions from ArviFoxy, @ZhongpaiGao, @mrowdays, Sil, timbalud, Devin Watson, Kari Roegnvaldsson, nielstron, Akash Levy, Tim Gehrunger, Baiying Liu, Argonaut Math, me, and OpenAI's original 7/8 preprint.
Our repo doesn't require the full Lean verification to submit, so it's easier to contribute: github.com/rohanarun/quasi-r…
I'll still run the full Lean verification in parallel for submissions.
We have new sponsor @Agent37HQ for QRH now, so the first 100 contributors also get free sandboxes to run experiments/searches!
It's like math Christmas 😂 We had so many improvements across multiple projects today.
github.com/rohanarun/quasi-r…
I made a catalog of math components (lemmas, identities, obstructions…) and Lean declarations/modules for improving the QRH bound.
It's open for collaboration. I have a script/agent monitoring for PRs & handling review (6.1 Sol with ability to escalate to Astra), so when my laptop is connected PRs will be approved quickly.
I really want to see this research move forward! Surely we can get as many collaborators as the integer multiplication folk?
github.com/HaileyStorm/QRH-C…
(Yes, the README is AI-written, pardon my laziness)
gtimg.github.io/QuasiRiemann…
haileystorm.github.io/resear…
Thank you @TimGMath for making this tracker, and thank you @ShivaKintali (again) for the short-proof inspiration!
My previous non-record paper, proving 4/33 in only 36 pages (also has Lean verification): teapaw.org/HaileyStormC/status/21…
c = 4/33 proven in 36 pages!
This inspired me. I set out, with GPT6-Pro of course, to find a short proof for a larger value of c. Mine isn't 27 pages, but it's a much larger half-plane!
Formalization locally verified in Lean, NanoDa, and con-ron.
PDF & formalization links in reply.
Thank you for the inspiration and starting point, Shiva!
More QRH news to come ;)
