Yichen Sheng

Introducing DLSS 4.5: ⚫️Second-gen Super Resolution transformer model for all RTX GPUs ⚫️Dynamic Multi Frame Gen for RTX 50 Series GPUs in Spring '26 ⚫️6X Multi Frame Gen for RTX 50 Series GPUs in Spring '26 ⚫️DLSS Overrides in NVIDIA app Learn More → nvda.ws/4aMEo5h

Computerbase did a blind video image quality test with thousands of votes and several games, and found a strong preference for DLSS 4.5 image quality at 2X upscaling over Native full resolution rendering. When we launched the Turing GPU generation, we knew the day would come when Neural Rendering even at a low resolution would be more detailed and better looking than traditional methods, while also being significantly faster. And we are just getting started! techpowerup.com/346494/blind-t…

🎉 **Scenethesis** has been accepted to #ICLR2026 ! Agentic systems are everywhere right now—coding agents, robotics agents, tool-using agents #moltbook. Back around two years ago when #OpenClaw , #NanoBananaPro have not arrived, we asked: can an agentic workflow *build simulation-ready 3D worlds* from text prompt? Interactive 3D scene generation isn’t just “generate some assets”. The hard part is spatial intelligence: ✅ spatial realism ✅ support & affordances ✅ physically plausible, editable, interactive layouts Scenethesis is a *training-free* language + vision agentic framework that: - 👁Let agents see: The planner doesn’t operate blind: it gets visual feedback and can correct itself. - 🏠Go outdoor: Works beyond indoor rooms; handles more open, outdoor-style compositions. check our work at: Paper: arxiv.org/abs/2505.02836 Project page: research.nvidia.com/labs/dir/scene… #ICLR2026 #Agents #GENERATION #physicalai #spatialintelligence

Limitations & Discussion I-Scene achieves better layout than SAM-3D. 🔧 while instance quality can be futher improved: • Instance upscaling networks • Higher-resolution voxel spaces 🚀 Good news: #TRELLIS2 supports much higher resolution (1024³ vs. our 64³). 🌱 Takeaway: We identify a cheap, scalable path to training generalizable 3D scene models — a step toward immersive world-generation foundation models. 🧵 [6/6]

If you work in #3D generation or #worldmodel, definitely take a look at this: I-Scene is an image to 3D scene model, achieves better 3D scene-gen than #SAM3D just using limited amount of random dataset.

Do we really need massive curated 3D scene data for interactive world generation? #SAM3D, #WorldGen say yes. We say no. I-Scene learns better spatial knowlesge using only 25K randomly composed instances. 🔑 Key insight: We reprogram the instance generator to infer support, proximity, and symmetry from purely geometric cues for generating interactive scenes. 🧠 Scene-context attention 👁️ View-centric space 🧱 Random composition beats expensive curation 🌐 luling06.github.io/I-Scene-projec… 💻 github.com/LuLing06/I-Sce… 🧵 Details below [1/6]

🚀 NVIDIA hiring Research Scientist (all levels). GenAI for graphics/gaming: neural rendering, world models, real-time generation, AI characters. If you like turning research into products like DLSS used by millions, great fit. Apply: nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAEx… #Hiring #GenerativeAI