Ani Aggarwal

Vision AI researcher |
Applying for 2026 PhD |
CS + Math from UMD |
I like vision

[Ani Aggarwal @AnirudAgg]

🧵 Your DiT, faster
Introducing ECAD: we reframe diffusion model caching as multi-objective optimization and evolve Pareto-optimal schedules via a genetic algorithm—achieving 4.47 FID gain at 2.58× speedup, with no retraining or tuning.
🔗 https://t.co/RBF5SuguTd
#MachineLearning

[Ani Aggarwal @AnirudAgg]

Super excited to present our work on efficient upsampling of ViT, VAE, and any other latent features!

[Matthew Walmer @MatthewWalmer]

We’re looking forward to presenting UPLiFT at #CVPR2026! Efficiently extract pixel-dense features from pretrained backbones like DINOv3.

We’ll be at the final poster session on Sunday (6/7) from 3:30-5:30pm at Poster 474, so please come by!

Website: https://t.co/Wng5ZfzjGA

[Ani Aggarwal @AnirudAgg]

@heynavtoor Don't forget to mention his terminal emulator, kitty!!

[Ani Aggarwal @AnirudAgg]

@LunjunZhang Agreed on this. I'm very bullish on using EAs in particular in areas that gradients are expensive / impossible to commute and RL would be slow or unstable (see my paper ECAD). Combining them is super clever :)

[Ani Aggarwal @AnirudAgg]

@wimmer_th Congrats! Would love to chat in Rio!

[Ani Aggarwal @AnirudAgg]

🦖 Pixel 🦕 dense 🦖 DINO 🦕 features and image super resolution 🖼️!!
#CVPR2026

[Matthew Walmer @MatthewWalmer]

Excited to announce that UPLiFT has been accepted to #CVPR2026!

You can also try out UPLiFT right now to extract pixel-dense DINOv3 features with our pretrained models linked below!

Code: https://t.co/slgetOBNPO
Paper: https://t.co/9IqMewyZeG
Website: https://t.co/MJ78gJpXAJ

[Ani Aggarwal @AnirudAgg]

@massiviola01 Low memory mode is live!

[Ani Aggarwal @AnirudAgg]

@massiviola01 Yes! We have a low memory mode that is super efficient in memory, but adds some latency (by running some operations in sequence rather than in parallel). We will add this to the released code repo this weekend! Thanks for your interest in our work 😁

[Ani Aggarwal @AnirudAgg]

🎉 Thrilled to share that my first research paper has been accepted to #ICLR2026! 🎉
I’ll be attending in person in Rio and would love to connect with others! 🇧🇷

[Ani Aggarwal @AnirudAgg]

🧵 Your DiT, faster
Introducing ECAD: we reframe diffusion model caching as multi-objective optimization and evolve Pareto-optimal schedules via a genetic algorithm—achieving 4.47 FID gain at 2.58× speedup, with no retraining or tuning.
🔗 https://t.co/RBF5SuguTd
#MachineLearning

[Ani Aggarwal @AnirudAgg]

Technically more than one line of code 😁

[Ani Aggarwal @AnirudAgg]

Super excited to announce our paper! Upsample any latents (DINO, VAE, etc.) in linear time (compared to quadratic cross attention).
Our models are all available on Hugging Face and Torch Hub with just one line of code!
(please star the GitHub repo 🥺)
https://t.co/vHqlDxPTMe

[Matthew Walmer @MatthewWalmer]

We’re excited to announce UPLiFT, our lightweight, pixel-dense feature upsampler. UPLiFT boosts feature density, preserves semantics, and has better efficiency scaling than recent SOTA methods. See all links in the thread below.
Coauthors: @_sakshams_ @AnirudAgg @abhi2610
🧵[1/6]

[Ani Aggarwal @AnirudAgg]

@HrishbhDalal @MatthewWalmer @_sakshams_ @abhi2610 Agreed, but even without a larger size, I'm of the opinion that using a modern VAE and multi scale image generator for training (FLUX, maybe?, which has larger latents) would achieve really strong image super res, super efficiently.

[Matthew Walmer @MatthewWalmer]

We’re excited to announce UPLiFT, our lightweight, pixel-dense feature upsampler. UPLiFT boosts feature density, preserves semantics, and has better efficiency scaling than recent SOTA methods. See all links in the thread below.
Coauthors: @_sakshams_ @AnirudAgg @abhi2610
🧵[1/6]

[Ani Aggarwal @AnirudAgg]

@prodarhan This is sick, love the website side by sides and clean animations 😍

[Ani Aggarwal @AnirudAgg]

@QianqianWang5 Super interested and will definitely be applying! 😁

[Ani Aggarwal @AnirudAgg]

@prodarhan @IamKyros69 Only if you first reconceptualize diffusion caching as a multi-objective optimization problem and apply a genetic algorithm to find an optimal speed quality trade-off to do inference with

[Ani Aggarwal @AnirudAgg]

@LogitechG No thanks my secondhand G502 is going strong 7 years later

[Ani Aggarwal @AnirudAgg]

@prodarhan can we get a comparison to Mr Pilu as a baseline??

[Ani Aggarwal @AnirudAgg]

@psandovalsegura Super interesting! Also seeing computational redundancy in DiTs. Some tokens cache well across timesteps (arXiv:2410.05317) and some entire blocks can be skipped (hasty plot). Maybe they could be pruned like your attention heads? Though uncertain of its practical application haha