Nima Fathi

Research Scientist @nvidia | ex @ServiceNowRSRCH, @mila_quebec , @shopify

[Nima Fathi @Nima0Fathi]

@mariannearr @nvidia Thank Marianne!

[Nima Fathi @Nima0Fathi]

Still feels surreal to say, but I’ve officially started as a Research Scientist at @nvidia, working on pretraining diffusion language models. 🚀

A few years ago, this was genuinely a dream. Grateful for everyone who supported me along the way.

[Nima Fathi @Nima0Fathi]

Presenting “Unifying Autoregressive and Diffusion Language Models” at #COLM2025 today! 🚀
Come chat if you’re curious about bridging AR & diffusion models (Poster #48, 16:30–18:30).

Project Page: https://t.co/D7yPhkGokB
ServiceNow Blog: https://t.co/bdsPVHtVpm

#dLLM #dLLMs

[ServiceNow AI Research @ServiceNowRSRCH]

🚀 New Research Blog Live!

Our latest post is out: Unifying autoregressive & diffusion language models by Nima Fathi, Torsten Scholak, and Pierre-André Noël.

𝗔𝘂𝘁𝗼𝗿𝗲𝗴𝗿𝗲𝘀𝘀𝗶𝘃𝗲 and 𝗱𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹𝘀 have each driven major advances in generative AI — but until now, they’ve been studied mostly in isolation. In this blog, our ServiceNow Research team explores a unified view that brings together 𝘁𝗵𝗲 𝘀𝘁𝗿𝗲𝗻𝗴𝘁𝗵𝘀 𝗼𝗳 𝗯𝗼𝘁𝗵 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗲𝘀.

🔍 Inside:
• A perspective unifying autoregressive and diffusion models
• How this unification could lead to more flexible and powerful language models
• Novel hybrid discrete diffusion processes teaching the model how to fix its own mistakes
• Early findings and research directions for the field

👉 Read more: https://t.co/pIE5joaERO
⚡ Website: https://t.co/JCBHH64HcT

#AIResearch #LanguageModels #DiffusionModels #Autoregressive #GenerativeAI

[Probabilistic Vision Group @PVG_McGill]

🚀 Our paper DeCoDEx has been shortlisted for Best ORAL paper at #MIDL2024! 🎉 Congratulations to PVG students @Nima0fathi and @amar_kumar_1 for this great achievement!

📄 Paper: https://t.co/A9kmFYNzsD
💻 Code: https://t.co/fgCMFie2KV
📹 YouTube: https://t.co/3fM7Pcs1CU

[Nima Fathi @Nima0Fathi]

Excited to present on stage in just 30 minutes! Watch our presentation live here: https://t.co/tEgBn2lyYU
@MIDL2024 #MIDL2024

[Probabilistic Vision Group @PVG_McGill]

Join us for the oral session @MIDL2024 on "DeCoDEx: Confounder Detector Guidance for Improved Diffusion based Counterfactual Explanation." Happening in 30 mins.

Streaming link: https://t.co/nhVuJSF1MR

Paper: https://t.co/ivmvX7TABy

[Nima Fathi @Nima0Fathi]

So thrilled to share my latest work accepted to #MIDL2024! Check out the full thread for more details on our DeCoDEx framework for improving counterfactual explanations in medical imaging.

[Probabilistic Vision Group @PVG_McGill]

Excited to share our latest work on debiased counterfactual explanations in medical imaging! Our DeCoDEx framework improves diffusion-based counterfactuals using a pre-trained confounder detector. #MIDL2024 @PVG_McGill @McGill_CIM @Mila_Quebec @GoogleAI 👉 https://t.co/NRHj80m9VL