Top Tweets for #diffusions
My thoughts on #MachineUnlearning research (Part 2): are our generative models (#LLMs and #Diffusions) actually forgetting data, or are they just breaking?
While classifier unlearning is an important starting point, the real-world stakes are much higher for LLMs and diffusion models. Generative AI faces real copyright and privacy issues right now, and retraining foundation models from scratch is out of the question due to the expensive cost. Naturally, generative unlearning is exploding with new papers every week. But the race to claim SOTA on existing benchmarks has become even more extreme, while discussions on how to properly evaluate unlearning have taken a back seat.
Today, the community mostly evaluates unlearning with narrow proxies, like measuring likelihood on a few fixed answer templates or using an image classifier. So we introduced FADE to check if an unlearned model actually matches an oracle model trained from scratch. The results were eye-opening. SOTA methods that look great on paper actually drift further away from the retain oracle. To pass the benchmark, LLMs collapse into repeating meaningless words or spitting out gibberish, while diffusion models wash out their images into style-neutral mush. The benchmark says unlearning succeeded, but in reality, the model is simply broken.
When we submitted this paper to top conferences, the reviews were another classic experience. Questioning standard benchmarks was once again treated like touching a sacred cow. Some reviewers even agreed that our evaluation reveals the truth, but argued that running a proper evaluation against a retain oracle is just too computationally expensive compared to existing shortcuts. In other words, "your evaluation is right, but the wrong evaluation is cheaper, so let's stick with the cheap one."
Jokes aside, could anyone actually deploy these models in production? If unlearning just means breaking a model until it passes a narrow benchmark test, we are only fooling ourselves. Without better evaluation, unlearning risks becoming a leaderboard game with zero practical value. If you work on #LLM or #diffusion unlearning, what do you think? How can our community move past convenient benchmarks toward real unlearning?

Super excited to be here in NYC to continue working on all the things I love most (... #flows, #diffusions, #biophysics, oh my!) 😃
Looking forward to interacting with the community at @NYUDataScience and @FlatironInst
I think the flow matching/stochastic interpolant convention is more natural, especially when you look at these types of models (especially diffusion models) as latent variable models as we explain in SDE Matching
https://t.co/XDfueQcUpv
#GenAI #Diffusions #Flows #ML

Why did the flow-matching community settle on t=0 for the prior and t=1 for the clean data distribution? Curious about the folklore here
A new Patreon Tier 3 interview with Michael Eisenberg from Metapsychosis Journal @metapsychonaut on Acousmatic Music & the Mundus Imaginalis is now available on Patreon!
#acousmaticcrossings #esotericism #diffusions #mundusimaginalis #henrycorbin
https://t.co/yVxDrqfQgb
Looking forward to attend and give an invited talk on our recent diffusion model work at the Generative Models and Uncertainty Quantification workshop in Copenhagen next week!
https://t.co/pWOgZr3gbw
Really a great line-up of speakers.
#GenAI #UQ #ML #Diffusions #Flows
#AI art at #4k / #EVERTIDE 💥 NOW has Browser Version 💥 Advanced Stylistic Approaches and Adversarial #Diffusions surpassing #midjourney and #leonardo

'Mean-Square Analysis of Discretized Itô Diffusions for Heavy-tailed Sampling', by Ye He, Tyler Farghly, Krishnakumar Balasubramanian, Murat A. Erdogdu.
https://t.co/dxi6g9Tc36
#itô #complexity #diffusions
#AI art at #4k / #EVERTIDE Desktop software with Advanced Stylistic Approaches and Adversarial #Diffusions surpassing #midjourney and #leonardo

I was honored to write a preface for this book presenting #generativeai explorations by 22 architects #diffusions #artificialintelligence #generativeai #AI #aiarchitects #aiarchitecture #neuralarchitecture https://t.co/FsFjdZBjs1

#Mathematics Article by Mohamed E. Nasr (et al.)
Analysis of #Non-Linear Radiation and Activation Energy Analysis on Hydromagnetic Reiner–Philippoff #Fluid #Flow with Cattaneo–Christov Double #Diffusions
https://t.co/uZk8BuqIxq
@MDPIOpenAccess @ComSciMath_Mdpi

Many thanks to the organizers @GSummerSchool for putting together such an enriching 2 weeks delving into #StatisticalPhysics of #ComplexSystems #phaseTransitions, #diffusions #outofequilibrium #topologicaldefects. 🧠💡📚
Segona i darrera setmana de la School on #StatisticalPhysics of Complex Systems, amb més cursos i seminaris.
A la imatge, la chair Carmen Miguel, el ponent @josperello i foto de família amb els participants en la Facultat de Física de la @UniBarcelona
🖱️https://t.co/hsCqKyds9J

#Diffusions 🔥
Ne manquez pas les matchs à l'affiche, tous les soirs dès 20h sur @lafaillesauvage !
📺 https://t.co/8tW3Q1ygX9

Exclusive: Interpol Publishes New Insight on Red Notice and Diffusion Screening Processes
@rogersahota summarises the key points from #Interpol's unexpected new information on the mechanisms used to screen and block abusive #RedNotices and #Diffusions https://t.co/uchkPc96Z4
Soft Diffusion: Score Matching with General Corruptions
Giannis Daras, Mauricio Delbracio, Hossein Talebi, Alex Dimakis, Peyman Milanfar.
Action editor: Jonathan Scarlett.
https://t.co/CoRKyOEoLm
#denoising #diffusion #diffusions
Soft Diffusion: Score Matching with General Corruptions
https://t.co/VzbUFh5ypp
#denoising #diffusion #diffusions

! #Eyekard vous souhaite une bonne rentrée !
🍀Voici, quelques souvenirs à partager de 21/22 de #diffusions + collaborations géniaux.
🍀MERCI @Halle_LaMachine @QuaiDesSavoirs @theatrecapitole @TlseMetropole @clubdelacom_ @DareWin @museumtoulouse @CouventJacobins @QualivoresOc

#Merci à toutes les radios qui diffusent mon titre #Futile 🙏🏼
Ce matin, petit focus sur :
• @PepsupMusic et @rvrradio qui ont intégré mon single à leur playlist
• @LaRvvs qui fait entrer mon titre demain en rotation.
#single #frenchpopmusic #poprock #radio #diffusions

Researchers from @TsinghuaSIGS analyzed the macroscopic and microscopic #diffusions of nuclear pollutants in the ocean. The results are important for quantitative prediction of long-term #radionuclide diffusion. For more: https://t.co/zrEr0QvJGl

La Bretagne, ça vous gagne 🎶🎵
Merci à @rfzbretagne qui diffuse #Futile 🙏🏼
#radio #bretagne #music #landerneau #diffusions #diffusion #frenchpopmusic #poprock #single

A votre avis, qu’est-ce qu’il se passe en face de Rose Bellecour ? 😱🔪
Réponse le vendredi 10 décembre à 21h sur @France2tv
#petitsmeurtres #lespetitsmeurtresdagathachristie #saison3 #serie #france2 #agathachristie #arthurdupont #emiliegavoiskahn #chloéchaudoye #diffusions

Last Seen Hashtags on Sotwe
bennybbc()*
Seen from New Zealand
benikiskandiramazsin
Seen from United States
อาบน้ำกันๆ
Seen from Thailand
重口
รับงานจันทบุรี
Seen from Thailand
arab edit
Seen from Turkey
anxiouspanda video ()*****************
Seen from United Kingdom
臉書限動原片
Seen from Thailand
मुद्दा_गरम_है
Seen from United States
AngelaWhite
Seen from Singapore
Trends for you
Most Popular Users

Elon Musk 
@elonmusk
240.5M followers

Barack Obama 
@barackobama
118.7M followers

Cristiano Ronaldo 
@cristiano
114.4M followers

Donald J. Trump 
@realdonaldtrump
111.8M followers

Narendra Modi 
@narendramodi
107M followers

Rihanna 
@rihanna
98.7M followers

NASA 
@nasa
92.2M followers

Justin Bieber 
@justinbieber
91.8M followers

KATY PERRY 
@katyperry
90M followers

Taylor Swift 
@taylorswift13
84M followers

Lady Gaga 
@ladygaga
75.5M followers

Virat Kohli 
@imvkohli
73.4M followers

Kim Kardashian 
@kimkardashian
70.8M followers

YouTube 
@youtube
68.7M followers

Neymar Jr 
@neymarjr
66.3M followers

Bill Gates 
@billgates
65.2M followers

Selena Gomez 
@selenagomez
63M followers

The Ellen Show
@theellenshow
62.1M followers

CNN 
@cnn
61.7M followers

X 
@x
60.6M followers































