Imran Ullah

I create UGC-style short video ads for products and online brands. Product demos, testimonials, hooks, captions, voiceover, and clean editing for TikTok, Reels, Shorts, and paid social ads. Paid work only. DM me with your product, budget, and timeline. #UGC #UGCAds #UGCCreator

@ash_csx Who write this post? 🤣

We’re dropping two open source SLMs this week. 1. One of them matches SOTA accuracy at up to 93x smaller. 2. The other one beats a recent OpenAI model. Model #1 drops tomorrow 👀

@saltjsx If the throughput is high, so the api cost should be lower.

Introducing MOG-1, the world's most powerful model. MOG-1 excels at deep reasoning, agentic coding, and advanced problem solving. It scores higher than any other publicly available model.

@ilijabogunovic Interested in remote opportunity.

Hiring 2 summer ML research interns at the University of Basel 🇨🇭. Research topics: RL/diffusion LLM post-training, reasoning, or LLM orchestration. Possible fully funded PhD offers to follow. I'll be at ICLR this week and happy to chat. Apply: forms.gle/TeeeNU6e7kDH3j…

@jeankaddour You are using gradient clipping that's why the loss is look like steady.

ML interview question: What is happening here?

I was rejected again from Erasmus Mundus IPCVAI and DEAI. This was my second attempt. Disappointed, but still committed to my journey in AI. I will continue learning and improving. #Erasmus #DEAI

@bansalg_ Interested

We're hiring. Over the past year, our team shipped: - AutoGen (now Microsoft Agent Framework) - Magentic-One - Magentic-UI - Magentic-Marketplace - MarkItDown - Fara-7B If you want to build AI agents that actually work *with* people—not replace them—come join us. DM me or @SaleemaAmershi

@YueYangAI @allen_ai I am interested, but the application form is too long to fill.

Our PRIOR team @allen_ai is recruiting research interns for Summer 2026! Topics include Language & Vision, Embodied AI, Agents, and more. If you are excited about VLM robustness, synthetic data, and vision-based agents, please select me as a potential mentor in your application. I will also be at NeurIPS from Dec 4–5. Apply here: job-boards.greenhouse.io/thealleninstit…

@jlambert_ Submit the Google form.

If so, we want to hear from you! To apply, please follow these two steps: 1️⃣ Fill out the form below to personally express your interest to me: docs.google.com/forms/d/e/1FAI… 2️⃣ Fill out the Google DeepMind "Student Researcher 2026" application form: google.com/about/careers/…

I am hiring a Student researcher at Google DeepMind for Winter/Spring or Summer of 2026! 🛠️ Interested in tool-use and reasoning for LLMs? 🚀 Up-to-date with frontier research? 🔬 Love to understand code & algorithms from first principles? 📚 Currently studying for a BS/MS/PhD?

@elonmuskTN 11

@Ai2Grinda82233 🔥🙃

Research Paper Roundup 🔬 18 breakthrough AI papers just dropped that challenge 40 years of assumptions! Key highlights: - New training methods beat backpropagation with 41% less energy - AI learns scientific truths humans can't observe - Longer reasoning ≠ smarter AI (surprising!) - Local learning rules > global optimization - The future of AI is being rewritten. Right now. 📚⚡ #AIResearch #MachineLearning #Breakthrough

@Ai2Grinda82233 Awesome

Mercury Diffusion Model Breakthrough Mercury just shattered the speed barrier! The world's first commercial diffusion language model generates text 10x faster than traditional LLMs - over 1,100 tokens/sec on H100s! Unlike sequential models, Mercury refines entire outputs like image AI, enabling parallel token generation. This isn't just faster - it's a paradigm shift! 🔥 arxiv.org/pdf/2506.17298 #AI #Mercury #DiffusionLLM #TechBreakthrough

Do you remember when you joined X? I do! #MyXAnniversary

@NandoDF If the data quality is good then 10k is enough for SFT, and for RL you need to collect or generate at least 1k-3k sample.

If you have 10K data instances, would you: 1. SFT an LLM with 10K data, or 2. Learn a reward with 5K, and RL the LLM on the remaining 5K with the learned reward 3. Other (explain)?

@iScienceLuvr Ms-swift framework

RLVR/RLHF libraries: • verl - ByteDance • TRL - HuggingFace • slime - Zhipu AI • prime-rl - Prime Intellect • ROLL - Alibaba • Nemo-RL - NVIDIA • AReaL - Ant Research • SkyRL - UC Berkeley • open-instruct - Allen AI • torchtune - PyTorch Any I am missing? Which do you like?

@shawnshenjx Nice work.

I’m Shawn, founder of Memories.ai, former researcher at Meta and CS PhD at University of Cambridge. Today we’re launching Memories.ai: we built the world’s first Large Visual Memory Model - to give AI human-like visual memories. Why visual memory? AI to date is chat-based, which has great applications, but humans are not only chat based. Humans have visual-memories, and are visually driven. Memories.ai enables AI agents, softwares, and even robots to see and remember the world the way humans do. We just raised an $8m seed round led by @susaventures to build the visual-memory layer for AI. Why is Memories.ai so groundbreaking? 👇

@_vztu 😀😂

Dear NeurIPS reviewers, please be reminded to delete the GPT prompts next time :)

I'm facing training instability issues while fine-tuning Qwen3-14B with Unsloth. Dataset size is 1M. I reduced the learning rate from 3e-4 to 2e-4, but sudden spikes in grad-norm. How can I fix this?

@mdancho84 This one still not working for excel file with embedded images.

🚨 BREAKING: IBM launches a free Python library that converts ANY document to data Introducing Docling. Here's what you need to know: 🧵

Accelerating Long-Sequence Transformers with Ring vs. Striped Attention on Multiple GPUs @imranullahds/accelerating-long-sequence-transformers-with-ring-vs-striped-attention-on-multiple-gpus-4615da572af1" target="_blank" rel="nofollow noopener">medium.com/@imranullahds/…