david yan

phd student @PrincetonVL, research intern @elorianai.
formerly: bse @princetoncs

[david yan @dzyan01]

@lukasmunzel congrats Lukas!

[david yan @dzyan01]

Huge thanks to all my collaborators on this paper, @Le_Xue01, @forrestnz, @richard_qiuyi_z, @dustinvtran, @AndrewDai, and @yinfeiy! Work done as part of my internship at @ElorianAI. 7/7

Project page: https://t.co/caQ9HzYlTK
BrickVision: https://t.co/Fkck9gLfQT
Code: https://t.co/AzrRfefK2J

[david yan @dzyan01]

Grounded reasoning is a popular thinking strategy for VLMs. But do models even use their grounding? (spoiler: not always!)

In our @ElorianAI paper, we dig into the grounded reasoning recipe and release BrickVision, a benchmark that’s challenging to solve without grounding. 1/7

[david yan @dzyan01]

On BrickVision, existing models vary widely in performance! 6/7

[david yan @dzyan01]

@huksyn Really cool result! We have some really high-fidelity + super fast export2sim stuff coming soon to Infinigen to help make this kind of work even more scalable :)

[david yan @dzyan01]

super cool use of infinigen for zero-shot sim2real!

[Siyuan Hu @huksyn]

We just released HALO — Perceptive Humanoid Motion from Simulation.

Trained entirely on synthetic data from 1,000 procedurally generated homes. Zero-shot deployment in 6 unseen real environments. 91% success rate across 360 physical trials. Fully onboard.

Paper, code, and project page are all live.

https://t.co/ZqFakAxmXT

[david yan @dzyan01]

@johnsmithyson0 Don't disagree that strict impact/progress can be faster in industry! It just means that in order to get positive net reward from doing a PhD, one will need to have high personal value on things such as research freedom, working in public as opposed to secrecy, etc.

[david yan @dzyan01]

Random thought on "why do a (ml/ai) phd?" I believe AI has shortened one's optimization horizon, due to high future uncertainty. As such, you should be confident that you will get positive short-term value, instead of indexing on the eventual delayed reward from finishing.

[david yan @dzyan01]

"Personal reward" here refers to some abstract net reward relative to other options (e.g. working in industry). Person b is not necessarily unhappy or sad, but just could be happier elsewhere! In contrast, doing a PhD is locally optimal for person a.

[david yan @dzyan01]

@wenhaocha1 everyone purchased the same weight-editing RL environment lol

[david yan @dzyan01]

@PrincetonVL @jiadeng To reproduce the demo scene, just run

python -m infinigen_examples.generate_nature --seed 2 --task coarse populate fine_terrain --output_folder out/mountain_2 --configs mountain noisy_video performance -p "execute_tasks.frame_range=[1,10]" land_process_pool.enabled=True

3/3

[david yan @dzyan01]

Excited to announce a major performance update for Infinigen! We've overhauled various assets and systems, focusing on efficiency. Our `--performance` config reduces the average generation time of nature scenes by 8x without sacrificing any visual fidelity! 1/3

[david yan @dzyan01]

All scenes are generated from fully procedural code, so no external assets are required. We hope you (or your agent) will give it a try at https://t.co/cMViWwUjn7! For more updates, follow @princetonvl. This performance update was led by @jiadeng and I. 2/3

[david yan @dzyan01]

my bitter lesson is finding out that caffeine significantly increases my focus and productivity

[Kartik Garg @_pacificax]

Sprout Jackson

[david yan @dzyan01]

@ChongZzZhang very happy it's independent of isaac now!