arxiv.org/abs/2504.07491 Kimi-VL’s technical report is out on arxiv!

Have you seen RecML library from Google?github.com/AI-Hypercomput… Eugene Yan

Weird thing happened. Gemini 2.5 bot in Windsurf Current wanted to access another unrelated repo in Cursor WTF! Google Gemini App

Önce Rusya, sonrasında İran… Yetişmiş, ülkesine bağlı insan gücünü kovalamış, yitirmiş, herkesi “işbirlikçi/komprador” gören, kendi dar çıkar çevresine sıkışmış, halkı sefalete sürüklenen, fukaralığın propagandası yapılan ülkelerde nükleer mükleer işe yaramıyor.

I have not used API yet but I am trying it on web app (gemini.google.com/app) and windsurf/cursor so no thinking budget. As a side effect, model's prompt adherence or instruction following also deteriorate. Just as an example: when I prompted for PRD with earlier version, it

Wrapped up Stanford CS336 (Language Models from Scratch), taught with an amazing team Tatsunori Hashimoto Marcel Rød Neil Band Rohith Kuditipudi. Researchers are becoming detached from the technical details of how LMs work. In CS336, we try to fix that by having students build everything:

Multi-vector search is beating agentic RAG by miles. But they come at a cost. Or do they? Recently, new multi-vector models like Reason-ModernColBERT have outperformed popular reasoning-based retrieval models like ReasonIR-8B. The embedding-per-token approach of ColBERT style

RAG is the most critical part of context management in AI. But doing it right is tough. I created a free, interactive simulator that visualizes different variants: 🧵

🤔 Have you ever wondered how good ModernBERT is compared to decoders like Llama? We made an open-data version of ModernBERT and used the same recipe for encoders and decoders. Turns out, our encoder model beat ModernBERT and our decoder model beats Llama 3.2 / SmolLM2 🤯 🧵

'water is transparent only within a very narrow band of the electromagnetic spectrum, so living organisms evolved sensitivity to that band, and that's what we now call "visible light". ' (found via HN)

So before people take credit, I found the oai os a min after they uploaded and saved the config and other stuff before it was removed. It’s an OS model and coming soon so kinda feels like ruining a surprise

Our open models are here. Both of them. openai.com/open-models

Today's library addition: A new (2024) history book on how we arrived at the current notation for vectors and general tensors, with an early nod to the quaternion approach and how things could have gone another route.

Presenting the GLM-4.5 technical report!👇 arxiv.org/abs/2508.06471 This work demonstrates how we developed models that excel at reasoning, coding, and agentic tasks through a unique, multi-stage training paradigm. Key innovations include expert model iteration with

GEPA is here! Thank you DSPy github.com/stanfordnlp/ds…

The percentage of child dead in the Ukrainian conflict is 0.3%. Zero point three per cent. The percentage of child dead in the Gaza conflict is 37.7%. Thirty seven point seven per cent. In a single statistic, there you have the difference between a war and a genocide.

We open-source Ring-flash-2.0 — the thinking version of Ling-flash-2.0. --> SOTA reasoning in math, code, logic & beyond. --> 100B-A6B, 200+ tok/s on 4×H20 GPUs. --> Powered by "icepop"🧊, solving RL instability in MoE LLMs.