Luigi Sigillo

Postdoctoral Fellow @ EMBL | Previously PhD in ICT @SapienzaRoma | Generative AI | |

[Luigi Sigillo @luigi_sigillo]

@peterottsjo Great thread!
Many of the systems you mention ultimately depend on the biomedical literature.

Today we released EMBL AI LIBRARIAN, a knowledge layer that helps life-science AI agents access that literature through natural-language queries.

https://t.co/JfuKQFiL6P

[Luigi Sigillo @luigi_sigillo]

@Silver22_2k @erpasd @JulioSaezRod @ola_spjuth @Fabio_Petroni

[Luigi Sigillo @luigi_sigillo]

🧬 Introducing 𝐄𝐌𝐁𝐋 𝐀𝐈 𝐋𝐈𝐁𝐑𝐀𝐑𝐈𝐀𝐍.
A knowledge layer for life-science AI agents.
Natural-language question in. Evidence snippet out.
No vector database. No separate literature index. Always up to date.
Open source ↓

[Luigi Sigillo @luigi_sigillo]

📄Research Paper:
https://t.co/JfuKQFiL6P

[Alessio Devoto @devoto_alessio]

Introducing NVIDIA Object Oriented Agents (NOOA), probably the most Pythonic way to build AI agents 🐍

In NOOA, an agent is just a Python class.

If you know Python, you already know NOOA. And yes, building agents can be this simple ↓

[Luigi Sigillo @luigi_sigillo]

@catalinmpit https://t.co/gbwgiC9hf0

[Luigi Sigillo @luigi_sigillo]

@rajabi2001 Excellent work! 👏 It’s exciting to see others tackling the high-res scaling crisis through frequency analysis. Our ICLR '26 paper on Latent Wavelet Diffusion follows a very similar intuition regarding spectral fidelity during denoising. Check it out:
https://t.co/l4jirF71b4

[Luigi Sigillo @luigi_sigillo]

@sedielem Excellent work! 👏 It’s exciting to see others tackling the high-res scaling crisis through frequency analysis. Our ICLR '26 paper on Latent Wavelet Diffusion follows a very similar intuition regarding spectral fidelity during denoising. Check it out:
https://t.co/vc5PnZqdGz

[Luigi Sigillo @luigi_sigillo]

@GordonWetzstein Cool work, congratulations! Also the intuition behind is similar to what we have done in our ICLR 26 paper!
https://t.co/vc5PnZqdGz

[Luigi Sigillo @luigi_sigillo]

@jo_schb Cool work, congratulations! Also the intuition behind is similar to what we have done in our ICLR 26 paper
https://t.co/vc5PnZqdGz

[Luigi Sigillo @luigi_sigillo]

📊 Project page: https://t.co/l4jirF71b4

🔗 OpenReview: https://t.co/vc5PnZqdGz

See you in Pavilion 4! Looking forward to discussions on ultra high-res synthesis 🇧🇷

[Luigi Sigillo @luigi_sigillo]

🇧🇷 Rio-bound for #ICLR2026

LWD wavelet trick unlocks razor-sharp 2K-4K images🔥

🔹 Wavelet energy maps create a smart frequency-aware mask
🔹 Detail-rich regions get extra training supervision

w/ @ShengfengHe @dhan90001
📆 Sat Apr 25 • 3:15-5:45 PM • Pav 4 • P4-#3207

[Danilo Comminiello @dhan90001]

2️⃣ 𝐋𝐚𝐭𝐞𝐧𝐭 𝐖𝐚𝐯𝐞𝐥𝐞𝐭 𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 𝐅𝐨𝐫 𝐔𝐥𝐭𝐫𝐚 𝐇𝐢𝐠𝐡-𝐑𝐞𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧 𝐈𝐦𝐚𝐠𝐞 𝐒𝐲𝐧𝐭𝐡𝐞𝐬𝐢𝐬
📆Sat, Apr 25, 2026 3:15 - 5:45 PM Pavilion 4 P4-#3207
👀 Site: https://t.co/iS98KhAZLF
📃 Paper: https://t.co/C6fBw1tZo5
w/ @luigi_sigillo @ShengfengHe

[Luigi Sigillo @luigi_sigillo]

This is my final PhD work, what a way to finish! 🎓Thanks to my supervisors @ShengfengHe and @dhan90001 .
Code 🔜
See you in Rio! 🌴

[Luigi Sigillo @luigi_sigillo]

🎉 "Latent Wavelet Diffusion For Ultra High-Resolution Image Synthesis" is accepted at #ICLR2026 🇧🇷!

💡Wavelet-based frequency masking to enhance 4K image generation 👉better details, zero inference overhead, no architectural changes needed.

📌Paper: https://t.co/vc5PnZqdGz

[Luigi Sigillo @luigi_sigillo]

@sainingxie To obtain better and stronger representations you can check our ICLR 2025 paper https://t.co/p64e6A4W9s

[Alessio Devoto @devoto_alessio]

Expected Attention compresses the KV Cache by estimating the attention score from future queries! Both during prefilling & decoding 🚀 All code released in our KVPress library & Leaderboard!

[Danilo Comminiello @dhan90001]

Yesterday we celebrated the PhD defense of @luigi_sigillo and @bong_taloma.
It was a wonderful journey full of professional but also brotherly experiences that I'll carry with me.
I couldn't be prouder of them, and I feel very lucky to have been part of their path🎊💙

@IspammL

[Eleonora Grassucci @eleonoragrassuc]

Three authors were not enough to present the poster to everyone at #ICLR25!😂

Forget the pairwise similarity. Enjoy the volume!

GRAM page👉 https://t.co/ORH4dz5si4

with @GiordanoCicchet @luigi_sigillo @dhan90001

#ICLR2025 #Multimodal