Introducing benchmark tracking across all of arXiv With thousands of models and benchmarks out there, it's impossible to track what's actually state-of-the-art We now index results from millions of papers to surface which models are SOTA according to open source research!

🚨 I am recruiting multiple PhD students this cycle to join my lab at Georgia Tech! If you're interested in problems at the intersection of robotics, machine learning, and systems for real-world human-robot collaboration, please apply! Application details in thread 🧵⬇️

A huge thank you to the hundreds who showed up to our panel last night on #AI4Science, with many audience members calling it their highlight of @NeurIPSConf! Special thanks to our panelists @jeffclune @james_y_zou @RGBLabMIT @bohannon_bot See our key takeaways below👇

Introducing RL Visualizer See PPO and GRPO mentioned everywhere but don't know what actually makes them different? Visualize and compare these algorithms in a simple online maze environment! 🚀

alphaXiv will be at NeurIPS in San Diego this week ✈️ Reach out if you'd like to chat and don't forget to check out our AI4Science Panel on Saturday (link below)

DeepSeek just announced DeepSeek-V3.2 2 open weights LLMs that go head-to-head with SoTA like GPT-5 high, Claude 4.5 Sonnet and Gemini 3.0 Pro With the highlight being their new DeepSeek Sparse Attention SoTA intelligence is now available for everyone to download

AI is transforming scientific discovery at incredible speed 🚀 Join us at NeurIPS for our AI4Science Panel with Rafael Gómez-Bombarelli, @jeffclune @james_y_zou as they discuss what the future of AI-enabled science might look like. Link to signup below 👇

This Meta paper just gave an empirical proof of why RL updates in LLMs are so sparse! They proposed this "Three-Gate Theory” & showed that pretrained models have highly structured optimization landscapes, so geometry-aware RL is much better than heuristic based SFT methods

Could evolution replace gradients for LLM training? Introducing EGGROLL, a new paper that makes gradient-free training actually practical 100x training throughput vs vanilla ES 🚀 100x memory reduction 💾 91% inference speed⚡ Pure integer training 🔢 Trending #1 on alphaXiv📈

AI research is accelerating at breakneck speed, and we need new platforms to support the dissemination of this info. @rajpalleti314 and the team at alphaXiv has been crushing it with impressive growth on their platform!!

Papers With Code SOTA is back 🚀 We used Gemini 3 to process millions of charts and tables across arXiv + the web to surface state-of-the-art AI research in reasoning, computer-use, OCR, and more Track the best-performing methods and see which benchmarks are getting adopted

Excited to see and support the next chapter of @askalphaxiv! Congrats @rajpalleti314 and team!

alphaXiv is truly transformational in how we interact with knowledge. It's been a real pleasure seeing the company grow as quickly as they have, a real testament to proactiveness of the team, especially Rehaan and Raj. Onto the better things they are going to achieve.

unbelievably proud of @rajpalleti314 and @rehaanahmad171, and grateful to have been able to advise such a brilliant and, most importantly, kind founding team. @askalphaxiv is going to reinvent how researchers discover and apply cutting-edge AI. this is only the beginning.

Huge congrats to the @askalphaxiv team! 🔥 There are few ideas that raise the collective outcome with no tradeoffs. alphaXiv has changed the ease of discovery of new ideas. A very worthy rebellion, alphaXiv is moving us beyond decades of sharing knowledge only using pdfs.

AI research volume has gone absolutely bonkers in the last 10 years. This product lets people keep up with and apply all of it. Today, millions of AI researchers come to AlphaXiv to see what the most important papers are, understand the core ideas, and even try hosted models

We just raised a $7M Seed round co-led by @MenloVentures and @haystackvc with participation from @Shakti_VC, @conviction and @upfrontvc 🚀 We're honored to have the support of incredible angels including @ericschmidt, @SebastianThrun, @sarahookr Join us: https://t.co/IKwK8KsG96

Diffusion LMs are more data efficient than Autoregressive (AR) LMs, and the difference is insane. Due to its training objective, DLMs basically have built-in Monte Carlo augmentation. So when unique data is limited, DLMs would always surpass AR models, and harder to overfit!

Introducing quickarXiv Papers are often written in convoluted language that is hard to understand. We're fixing that. Swap arxiv → quickarxiv on any paper URL to get an instant blog with figures, insights, and explanations. Now extracted with DeepSeek OCR 🚀

Hybrid linear attention that actually works 🔥 Kimi K2's Delta Attention matches full attention quality while cutting KV cache by 75% and achieving 6x speedup on 1M-token decoding We built an interactive playground + tutorial so you can test it live and build with it yourself