Yang Cao

曹洋 || CSE PhD student at @HKUST , supervised by Prof. Dan Xu @danxuhk || Interested in 3D vision & Multi-modal learning || Discussions are welcome :)

[Yang Cao @yangcao_]

Glad to share that our work, VGGT-Det (CVPR 2026). Existing multi-view indoor 3D detectors rely on precisely calibrated multi-view camera poses or depth—costly to obtain in real world. VGGT-Det changes this with a Sensor-Geometry-Free framework (no multi-view poses or depth)

[Yang Cao @yangcao_]

@XieYueqi Congrats! Great work!

[Yang Cao @yangcao_]

@JackyWangAI Thank you!😃

[Yang Cao @yangcao_]

Just got my first reviewer award. Glad to contribute to the community 😄

[Yang Cao @yangcao_]

Cool demo 👏

[Dan Xu ✈️ ICML'26 @danxuhk]

This project has been accepted to appear at ICML 2026 😇😇😇 @ICML #ICML2026 #ICML.

Our WildActor consistently preserves body identity (including facial features, body shape, and clothing details) under diverse shot compositions, large viewpoint transitions, and substantial motions.

Project page: https://t.co/JttAv6GsJY
Github code: https://t.co/IfOGFXdCWS
arXiv paper: https://t.co/tpzdz8CN2h

[Yuanchen Ju @ju_yuanchen]

Our #ICLR2026 Oral paper MomaGraph will be presented at Oral Session 3D Vision language models II, April 24th 10:30 BRT. Poster session is at Pavilion 3 P3-#1313, April 23rd 3:15-5:45 PM BRT.@furongh will be there, feel free to stop by and chat!💐

[Yang Cao @yangcao_]

@HongliZhan @jessyjli Congrats!!!

[Yang Cao @yangcao_]

@jihanyang13 @amilabs Congrats!!!

[PengtaoJiang @PengTaoJiang2]

We are excited to share our MagicWorld, a video world model that aims to address motion drift and long-horizon error accumulation. It achieves 20 FPS on the L40S GPU and better results on the RealWM120K-Val dataset under VBench. Project page: https://t.co/ha0v0R97Mb

[Elisabetta Fedele @efedele16]

Deadline for the proceedings track is today - last chance to submit your work! ⏰🤖

[Zedong (Jacky) Wang @JackyWangAI]

Thanks @HuggingPapers for sharing CARE-Edit! We introduce a condition-aware router that lets heterogeneous experts handle text, mask, and reference-guided edits without stepping on each other. Thrilled to see it accepted at #CVPR2026. Glad to have worked on this w/ @AndyYucheng.

[Yang Cao @yangcao_]

@Chain_AlphaX Hi, thanks for your interest. Please check our work here : https://t.co/YN0yGNgzWU

[Yang Cao @yangcao_]

Thank AK a lot for sharing our work 🤝

[AK @_akhaliq]

VGGT-Det

Mining VGGT Internal Priors for Sensor-Geometry-Free Multi-View Indoor 3D Object Detection

paper: https://t.co/fjQwdzF9ZW

[Yang Cao @yangcao_]

Code and pretrained models will be released at https://t.co/OcsZbYRVXE

[renjie pi @RenjiePi]

Introducing Nemotron-Terminal: a systematic data engineering pipeline for scaling LLM Terminal Agents.

We bridge the gap between open models and proprietary models with a fully open synthetic-to-real trajectory pipeline.

🤯The payoff: SFT on our Nemotron-Terminal-Corpus boosts Qwen3-32B from 3.4% → 27.4% on Terminal-Bench 2.0 (+24.0), rivaling models multiple its size.

What makes it work?

🌟Terminal-Task-Gen: A lightweight data curation pipeline that seamlessly combines the adaptation of existing datasets with robust synthetic task construction.
🌟Nemotron-Terminal-Corpus: A massive, open-source dataset covering diverse terminal interactions, which contains explicit planning and execution traces for complex long-horizon tasks.

And we’re releasing everything:

📦 Nemotron-Terminal-Corpus (Large-scale dataset)
🤖 Nemotron-Terminal models (8B, 14B, 32B)

Paper: https://t.co/ORIZ01sav1
HF Daily: https://t.co/nSH4hu7I5D
Models & Data: https://t.co/J1Zc22M95r

Our tech report just hit the #1 spot on Hugging Face Daily Papers!
We're also incredibly excited to see the open-source community putting our work to the test, with the Nemotron-Terminal-Corpus dataset currently trending at over 1,800 downloads and counting.
We can't wait to see what the community build with it!

[Elisabetta Fedele @efedele16]

🚨 Deadline extended to March 12 - you still have time to submit a proceedings paper to the OpenSUN3D workshop @CVPR !

If you are working on open-vocabulary 3D scene understanding, 3D perception, or spatial AI, we would love to see your work ✨🤖

➡️ https://t.co/4nSXaJGNpR

[Yang Cao @yangcao_]

On ScanNet and ARKitScenes, VGGT-Det outperforms the strongest SG-Free baseline by +4.4 and +8.6 [email protected], respectively.

Paper:
https://t.co/YN0yGNgzWU

Code (coming soon):
https://t.co/OcsZbYRVXE

[Yang Cao @yangcao_]

Glad to share that our work, VGGT-Det (CVPR 2026). Existing multi-view indoor 3D detectors rely on precisely calibrated multi-view camera poses or depth—costly to obtain in real world. VGGT-Det changes this with a Sensor-Geometry-Free framework (no multi-view poses or depth)

[Yang Cao @yangcao_]

VGGT-Det integrates the VGGT encoder into a transformer-based detector. With Attention-Guided Query Generation and Query-Driven Feature Aggregation, we unlock VGGT’s internal semantic and geometric priors end to end.

[Dan Xu ✈️ ICML'26 @danxuhk]

It has been accepted at CVPR 2026 #CVPR2026 @CVPR 😊😊😊

[Yang Cao @yangcao_]

👏👏

[Elisabetta Fedele @efedele16]

The submissions to our @CVPR workshop OpenSUN3D on Open-World 3D Scene Understanding and Representations are open! 🏔️🤖☀️

➡️ The deadline for the proceedings track is on the 5th of March!🏃‍♀️

🌍: https://t.co/4nSXaJGNpR
📝: https://t.co/TYZKnuKeSo