Scofield Hao Fei

🚀 Any-to-Any Multimodal Learning Workshop @ CVPR 2026 Exploring unified learning across vision, language, audio, 3D, video, and beyond. 📷 a2a-mml-2026.vercel.app 📷 Submission deadline: March 01, 2026 #Multimodal #AnyToAny #CVPR2026

The @karpathy interview 0:00:00 – AGI is still a decade away 0:30:33 – LLM cognitive deficits 0:40:53 – RL is terrible 0:50:26 – How do humans learn? 1:07:13 – AGI will blend into 2% GDP growth 1:18:24 – ASI 1:33:38 – Evolution of intelligence & culture 1:43:43 - Why self driving took so long 1:57:08 - Future of education Look up Dwarkesh Podcast on YouTube, Apple Podcasts, Spotify, etc. Enjoy!

@_akhaliq 🌍 UniVA is fully open sourced! ⚙ MCP-native modular design → plug&play new tools or APIs. 🚀 UniVA moves video AI from isolated tasks → true generalist intelligence. 👉 Try UniVA online demo system: univa.online 💻 Code: github.com/univa-agent/un…

🎥 UniVA is capable of industrial-grade video creation: 🔄 Any-conditioned (text, image, or video) controllable video 🎞 Cinematic quality with stable identity & object consistency 📽 Long-form, multi-scene, multi-shot narratives 📄 Paper: arxiv.org/abs/2511.08521

🎉 thrilled to introduce UniVA: Universal Video Agent towards Open-Source Next-Generation Video Generalist 🎥 An agentic, unified, omni-capable video AI, capable of planning, understanding, editing, and generating videos end-to-end 🌐 Website: univa.online

@_akhaliq [6/6] 💡 Our extensive evaluation experiment on 100+ popular existing MLLMs shows the great challenges in reaching genuine AI. Most MLLMs DO NOT achieve strong synergy, even hard to arrive at Level-4. Welcome submissions and contribution! 🏆 Leaderboard: generalist.top/leaderboard

@_akhaliq [5/6] ✨ To lower barrier, we design 4-scoped leaderboards: from full-spectrum highly capable MLLMs (Scope-A) to skill-specific partial generalists (Scope-D). Leaderboards will be run and maintained for long term. Go select a leaderboard best fiting your model’s strengths! 😎

On Path to Multimodal Generalist General-Level and General-Bench

Thx @_akhaliq for sharing our release: 🚀 General-Level, a synergy-centered 5-level generalist evaluation suite; 🍕 General-Bench, super massive multimodal benchmark. 🎉 Project: generalist.top 📝 Paper: arxiv.org/abs/2505.04620 🏆 Leaderboard: generalist.top/leaderboard

@_akhaliq [6/6] 💡 Our extensive evaluation experiment on 100+ popular existing MLLMs shows the great challenges in reaching genuine AI. Most MLLMs DO NOT achieve strong synergy, even hard to arrive at Level-4. Welcome submissions and contribution! 🏆 Leaderboard: generalist.top/leaderboard

@_akhaliq [5/6] ✨ To lower barrier, we design 4-scoped leaderboards: from full-spectrum highly capable MLLMs (Scope-A) to skill-specific partial generalists (Scope-D). Leaderboards will be run and maintained for long term. Go select a leaderboard best fiting your model’s strengths! 😎

[1/6] Excited to present our project: On Path to Multimodal Generalist: General-Level and General-Bench Accepted by ICML'25 (Spotlight).   🎉 Project: generalist.top 📝 Paper (300+ pages): arxiv.org/abs/2505.04620 🤗 HF Paper: huggingface.co/papers/2505.04…

Thanks @AK for sharing *JavisDiT*, a SoTA Joint Audio-Video Diffusion Transformer. 🤠 Our gradio demo is out, welcome trying out: 447c629bc8648ce599.gradio.live 📝 HF Paper: huggingface.co/papers/2503.23… ✨ Code: github.com/JavisDiT/Javis…