Junhyeong Cho

PhD student @Cornell @Cornell_Tech | multimodal spatial intelligence

[Junhyeong Cho @JunhyeongCho]

Where on the floorplan was this photo taken? 📍
Excited to share SceneAligner, accepted to #NeurIPS2026!
We tackle this by reconstructing the scene in 3D and aligning it to the floorplan. Joint work with @ruojin8 @ElorHadar [1/8]

[Hadar Averbuch-Elor @ElorHadar]

Excited to also share PersistBench that explores whether 4D foundation models remember (TL;DR not really!) #NeurIPS2026

[Junhyeong Cho @JunhyeongCho]

@ShamoonMaxxing @ruojin8 @ElorHadar Thank you!

[Junhyeong Cho @JunhyeongCho]

Where on the floorplan was this photo taken? 📍
Excited to share SceneAligner, accepted to #NeurIPS2026!
We tackle this by reconstructing the scene in 3D and aligning it to the floorplan. Joint work with @ruojin8 @ElorHadar [1/8]

[Junhyeong Cho @JunhyeongCho]

@AlexHe00880585 @Haozheng_Yu very cool!! 🤩🤩🤩

[Guangzhao He @AlexHe00880585]

Excited to share our #NeurIPS2026 E&D 🏆 Spotlight!

We all agree that memory is essential for modeling a persistent 4D world.
But: 🙋 Can 4D Foundation Models Remember?

Introducing 🔥 PersistBench: a benchmark for testing visual memory in camera-controllable video models / 4D reconstruction models / video-to-360° models.

One key challenge in evaluating visual memory is obtaining real-world ground truth for what happens to an object after it leaves the camera’s view.

In PersistBench, we introduce a simple way to sidestep this challenge. 🤫

🌐 Project Page: https://t.co/CjedC4pCMT
🏆 Leaderboard: https://t.co/r5lBmJkLEn
🐙 GitHub: https://t.co/g85Y5USjYP
🤗 Dataset: https://t.co/sSgCprESzH

See how well your models do and submit to our leaderboard today!

🧵 [1/n]

[Youwang Kim @youwangk]

I proudly share this new paper "𝗧𝗲𝘅𝘁𝘂𝗿𝗲 𝗦𝗽𝗮𝗰𝗲 𝗠𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻" from my NVIDIA mentors: Jacob Munkberg, Peter Kocsis (@Peter4AI), and Jon Hasselgren!

The idea is elegant: instead of generating material views in image space and fusing them, run diffusion directly in PBR material texture space. That makes the result view-consistent by design, while reusing strong pre-trained video diffusion priors to generate complete PBR materials that generalize across arbitrary geometries and UV parameterizations.

Honestly, the 8K-res PBR texture generation is the most impressive I've seen for this task. Enjoy the super-detailed lion statue in the attached video!

I'm also excited about its extension to neural materials, which connects beautifully to my own 𝙉𝙚𝙪𝙈𝙖𝙩𝙀𝙭 work with the same team on extracting neural materials from images. Two sides of the same story — pushing the boundaries of what we can do within PBR and with new representations beyond PBR.

Don't miss the stunning visuals on the project page!
🔗 Page: https://t.co/WGu6RQ1Ua3

Also check out the updated 𝙉𝙚𝙪𝙈𝙖𝙩𝙀𝙭 results 🤗
🔗 Page: https://t.co/LzYPpJgNo6

[Ziqi Ma @ziqi__ma]

Very excited to share LoGo, my internship project at @theworldlabs on post-training world models! We found that reward design matters a great deal in post-training long-horizon video gen for 3D consistency, echoing what we see in other domains, e.g. LLM reasoning. Check out more 👉 https://t.co/tLW6CVGaAk

[Junhyeong Cho @JunhyeongCho]

@YiwenZhangYZ Great work!! 🤩🤩🤩

[Ryohei Sasaki@engineer @rsasaki0109]

G3T Up! Gravity Aligned Coordinate Frames Simplify Pointmap Processing
Cornell University

TL;DR: We introduce G3T, a transformer that predicts upright, gravity-aligned pointmaps regardless of input image orientation, and G3T-Long, a pipeline that leverages this uprightness to enable robust long-sequence 3D reconstruction.

[Junhyeong Cho @JunhyeongCho]

@ElorHadar I've had so much fun working on this project!!! I'm so grateful for your guidance throughout my first PhD project 🙏

[Hadar Averbuch-Elor @ElorHadar]

Very excited about our #NeurIPS2026 papers, including SceneAligner that aligns a 3D reconstruction of in-the-wild images with the 2D floorplan map!

[Junhyeong Cho @JunhyeongCho]

@AlexHe00880585 So cool! 🤩🤩

[Guangzhao He @AlexHe00880585]

Ever suffered from making slides like I do? 😭

Aligning everything, fighting animations, trying to add cool interactive visualizations…while AI generated slides still look extremely confusing and overwhelming.

So I vibecoded JostSlides

⭐ https://t.co/8m995aAIPL

🧵 (1/n)

[Junhyeong Cho @JunhyeongCho]

@ShamoonMaxxing @ruojin8 @ElorHadar A few more examples:

[Junhyeong Cho @JunhyeongCho]

@ShamoonMaxxing @ruojin8 @ElorHadar Thanks! We tested on in-the-wild floorplans, including fairly complex layouts like this one 👇 We'd love to push this further to even more challenging floorplans!

[Junhyeong Cho @JunhyeongCho]

@jayfarei @ruojin8 @ElorHadar @huggingface Thanks! Yes, we're planning to release the model weights on @huggingface this week 😁

[Ruojin Cai @ruojin8]

✨ Really excited to see SceneAligner at #NeurIPS2026!

Love the cross-domain challenge: grounding images, 3D recons & floorplans in one spatial frame, even across scenes with little or no visual overlap.

Amazing work by Jun, and huge thanks to Hadar for all the guidance!

[Junhyeong Cho @JunhyeongCho]

@chaof1234 Congrats!! 🎉🎉🎉

[Junhyeong Cho @JunhyeongCho]

@lovvvyo3 @ruojin8 @ElorHadar Thank you! 😆

[Junhyeong Cho @JunhyeongCho]

@PulkitRega @ruojin8 @ElorHadar Thank you!!

[Junhyeong Cho @JunhyeongCho]

@ruojin8 Thank you, Ruojin! I really appreciate all your help. I had so much fun working on this together! 😆