HanRong YE
@leoyerrrr
Followers
892
Following
1K
Media
62
Statuses
371
@Nvidia Research Scientist | Open-Souce Omni-Modality LLMs
Hong Kong
Joined October 2012
Well our research finds that RL using FP4+LORA actually achieves faster reward growth and higher final accuracy than 16-bit LoRA and QLoRA: https://t.co/SmcEQMPiFv
BF16 -> FP16 is such a simple (one configuration change in Oat) yet fundamental fix for inference-training mismatch. With FP16, the most basic importance sampling PG outperforms all algorithmic fixes in BF16. Let's rethink RL stability from the precision perspective.🔎
0
0
4
📈 According to the official test results from OmniVideoBench using our open-sourced model weights, OmniVinci has once again claimed #1 performance in the 7B LLM category! In addition, the model has been out for just two weeks and downloads have already soared past 6K 🌟
2
0
4
560A27B omni-modal MoE has landed 🤗 …but OmniVinci-9B is still my favorite size 😃
🔥 LongCat-Flash-Omni: Multimodal + Low-Latency 🏆 Leading Performance among Open-Source Omni-modal Models ☎️ Real-time Spoken Interaction: Millisecond-level E2E latency 🕒 128K context + Supports > 8min real-time AV interaction 🎥 Multimodal I/O: Arbitrary Combination of
0
0
2
🚀Really excited to see this amazing arch change (KDA) finally coming out! Replacing global attention with linear hybrid arch: better pretraining ppls, long context evals, downstream math&code&stem evals after RL, >6 * throughput at 1M to unblock more downstream potentials to
Kimi Linear Tech Report is dropped! 🚀 https://t.co/LwNB2sQnzM Kimi Linear: A novel architecture that outperforms full attention with faster speeds and better performance—ready to serve as a drop-in replacement for full attention, featuring our open-sourced KDA kernels! Kimi
1
18
55
And we at #NVIDIA Research are still seeking research interns to explore omni-modal LLMs across a variety of domains, including robotics (VLA), visual agentic tool using, world modeling, and unified understanding and generation. Drop me an email if you are interested!
OmniVinci is now #1 paper on Huggingface!!! 🤗 Building omni-modal LLMs is MORE than just mixing tokens 😉 At @NVIDIA, we explored deeper possibilities in building truly omni-modal systems — leading to OmniVinci-9B, which introduces three key innovations: - OmniAlignNet – a
0
1
12
NVIDIA is stepping up our support of openly developed AI across the industry. https://t.co/j4dJ4UqDck
blogs.nvidia.com
NVIDIA’s open model families, including NVIDIA Nemotron for digital AI, Cosmos for physical AI, Isaac GR00T for robotics and Clara for biomedical AI, provide developers with the foundation to build...
4
13
49
Jensen is streaming live - AI, 6G, Quantum, Models, Enterprise, Robotics, Factories 💚 #NVIDIAGTC #NVIDIA
https://t.co/he1nst3nI9
0
1
6
🚀 📷🍌
🎁🎁 We release Pico-Banana-400K, a large-scale, high-quality image editing dataset distilled from Nana-Banana across 35 editing types. 🔗 Data link: https://t.co/mi06ddf3mN 🔗Paper link: https://t.co/AaZM02xcJr It includes 258K single-turn image editing data, 72K multi-turn
0
0
4
Your model can do COT reasoning — but is it actually correct?👿
🧩 New paper out! We introduce PRISM-Bench, a diagnostic benchmark for puzzle-based multimodal reasoning. Unlike standard VQA, PRISM-Bench tests not only if models can solve visual puzzles, but how their reasoning unfolds. 💡 Models must spot the first error in a
0
0
8
Several of my team members + myself are impacted by this layoff today. Welcome to connect :)
475
284
7K
@nvidia Joint work of @leoyerrrr , @yin_hongxu , @huckiyang , @goelarushi27 , @AaronWeiHuang , @LigengZhu , Yuanhang Su†, Sean Lin†, @anjjei , Zhen Wan†, @MXzBFhjFpS1jyMI , @YumingLou , Dong Yang†, @zhijianliu_ , @yukangchen_ , @AmbrishDantrey, @ehsanjjjjj , @SreyanG , Daguang Xu,
0
0
1
@nvidia 🔗 OmniVinci by NVIDIA Research 🌐 Webpage: https://t.co/qRscvzkSkh 💻 GitHub: https://t.co/2efUT4oS4C 🤖 Model: https://t.co/06nWW6gYO2 📄 Paper:
huggingface.co
1
1
6
Off to ICCV! Also, we have an omni-modal LLM reveal coming next Monday… straight from Hawaiiiii 🌴
1
4
34
Proud to see NVIDIA recognized by AI World as a leader in the open-source AI ecosystem. From Nemotron and BioNeMo to Cosmos, GR00T, and Canary, our contributions span foundation models, scientific computing, and agentic reasoning. I feel sooo excited to be part of the Nemotron
6
7
37
If you’re interested, please send your CV and cover letter to hanrongy@nvidia.com
0
0
1