Warock

Presenting AI-driven Voice Analysis in Esport 🔊: After a continued work of 6 months - 1 year with @PerezNicolasLol, we are proud to present some NLP - based techniques to construct metrics about voice communication effectiveness in Esport. - THREAD - 1/3

In an effort to try stop seeing so much slop I've been trying to train my own AI detection model. Found something incredibly interesting. for the most part LLM generated text and human written text are linearly separable.

@MagicaLyna Mais oui je comprend le fait que c’est de temps en temps exaspérant que ces mêmes personnes le crie sur tout le toit

@Warock42 Non mais biensur mais la je parle pas de ça

Faut voir le niveau de BRANLETTE intelectuelle des tiktoker League of Legends hein WOW La petite musique mystérieuse et les tournure de phrase qui explique comme si tu était un pro comment tu as win une lane en platine ou en emeraude franchement je m'incline devant un tel perf MDRRRRRRR

@MagicaLyna *pour ce genre de contenus

@MagicaLyna Si tu ne te complimente pas personne va le faire à ta place. Par contre c’est vrai qu’il y a un curseur à placer entre l'humilité et l'arrogance

Previous season for a professional team I was asked to assess whether grubs or drakes are more important for the expected outcome and the results where very interesting !

CHUCK NORRIS MEME THREAD TO END ALL MEME THREADS Let's honor one of the GOAT's by dropping your BEST Memes and Chuck Norris jokes! 👇👇👇

Putting out a wish to the universe. I need more compute, if I can get more I will make sure every machine from a small phone to a bootstrapped RTX 3090 node can run frontier intelligence fast with minimal intelligence loss. I have hit page 2 of huggingface, released 3 model family compressions and got GLM-4.7 on a MacBook huggingface.co/0xsero My beast just isn’t enough and I already spent 2k usd on renting GPUs on top of credits provided by Prime intellect and Hotaisle. ——— If you believe in what I do help me get this to Nvidia, maybe they will bless me with the pewter to keep making local AI more accessible 🙏

PyTorch interview practice platform github.com/duoan/TorchCode

wip

Introducing MCP for arXiv Let your research agents stand on the shoulders of giants Fast multi-turn retrieval, keyword search, and embedding search tools across millions of arXiv papers 🚀

Finally finished! If you're interested in an overview of recent methods in reinforcement learning for reasoning LLMs, check out this blog post: aweers.de/blog/2026/rl-f… It summarizes ten methods, tries to highlight differences and trends, and has a collection of open problems

Happy Pi Day! In a certain sense, π is not a constant, but a variable. Using our usual Euclidean distance, it is 3.14159… but applying other L^p norms on ℝ², half the unit circle's perimeter will give other values. For instance, at p=1 (taxicab geometry), “π” = 2√2. At p ≈ 2.2, it's 3.20. Anyway, the video I was hoping to have out this day will be out closer to the 20th. Some call it “missing your deadline”, but I prefer to think of it as giving the L_{2.2} norm a little love.

@dpmlol In the end it’s all gaussian representation

Role distribution among the Top 1000 players on the EUW ladder 📊

@iterogg Ornn Nocturne Orianna Ashe Seraphine.

What are your choices?

@l0v3milfs Elden Ring et Dark Souls existent :'(

comme j’ai la raison absolue je pense que vous avez tous tort, seuls outer wilds, nier automata et undertale peuvent être cités

@OnlyWaifu Little do they know, they’ve probably spent more hours than the so-called "tryhard"

People who call others "tryhard" just for trying to win are just coping with being dog water at the game

@drewlevin This should ve an interesting classification problem to tackle as the very nature of this behavior in the firdt hand is to mimick rhe nature of inting for fun. Hense the Kullback Leibler divergence of both behaviours is very small

ranked league players: should we treat flashing for fun (in the way that people do in aram) as a piece of evidence that someone is inting?

I've recently been struggling with the level of noise in LLM evaluations. Inspired by this, I did a deep dive into practical / applied statistics for LLM evals. Tomorrow morning, I'll publish my learnings so far in a long-form writeup on my blog. Statistics is a huge field, but a lot of the most important concepts for approaching evals in a rigorous manner are relatively easy to learn and apply. With a grasp of applied statistics for LLMx, you can: 1. better interpret results (i.e., understand if they are meaningful or caused by noise). 2. design evals in a way that is conducive to drawing more confident conclusions. Both of these points help us to run faster and more efficient experiments, rather than wasting time and compute chasing noise. Some of my favorite papers so far: - A statistical approach to LLM evaluations: arxiv.org/abs/2411.00640 - Don't use CLT in LLM evals with fewer than 100 data points: arxiv.org/abs/2503.01747 - Quantifying variance in evaluation benchmarks: arxiv.org/abs/2406.10229 - A framework for reducing uncertainty in LLM evaluation: arxiv.org/abs/2508.13144

We're hiring ~10 research engineers on the Interpretability team at Anthropic. If you're a seasoned ML infra engineer who's excited about understanding what's happening inside frontier models, we'd love to hear from you. (No prior interp experience needed!)