🤝 Great collaboration between @bocklab.bsky.social (@moritzbaio.bsky.social, Animesh, Jake), @nonchev.bsky.social, @gxxxr.bsky.social, and pathologist Viktor Kölzer.

SpotWhisperer is at #ICML25 FM4LS workshop. Visit our poster on Saturday (19 July 2025) if you're interested & attending ICML. (6/6)

Bluesky

🤝 Great collaboration between @bocklab.bsky.social (@moritzbaio.bsky.social, Animesh, Jake), @nonchev.bsky.social, @gxxxr.bsky.social, and pathologist Viktor Kölzer.

SpotWhisperer is at #ICML25 FM4LS workshop. Visit our poster on Saturday (19 July 2025) if you're interested & attending ICML. (6/6)

CompHealthCorner: What should you be reading this week? Check out this fantastic #ICML25 paper to learn how 'accurate' responses can be misinterpreted.

Position: We Need An Algorithmic Understanding of Generative AI What algorithms do LLMs actually learn and use to solve problems? Studies addressing this question are sparse, as research priorities are focused on improving performance through scale, leaving a...

📯 Come visit our #ICML25 Spotlight Poster and meet @taylorwwebb.bsky.social to discuss our work: "Toward an Algorithmic Evaluation and Understanding of Generative AI."

Paper: openreview.net/forum?id=eax...
Poster: icml.cc/media/Poster...

Presented with my co-supervisor @bokmangeorg.bsky.social at #ICML25, fun time!

I'm at ICML 2025!

DM me if you want to chat.

@icmlconf.bsky.social #ICML2025 #icml25 #BioNeMo

Accelerated Diffusion Models via Speculative Sampling, at #icml25 !

16:30 Tuesday July 15 poster E-3012

arxiv.org/abs/2501.05370

@vdebortoli.bsky.social Galashov @arnauddoucet.bsky.social

Distributional diffusion models with scoring rules at #icml25

Fewer, larger denoising steps using distributional losses!

Wednesday 11am poster E-1910

arxiv.org/pdf/2502.02483

@vdebortoli.bsky.social
Galashov Guntupalli Zhou
@sirbayes.bsky.social
@arnauddoucet.bsky.social

🤔 What are the properties of self-explanations in GNNs? What can we expect from them?

We investigate this in our #ICML25 paper.

Come to have a chat at poster session 5, Thu 17 11 am.

w. Sagar Malhotra @andreapasspr.bsky.social @looselycorrect.bsky.social

A great @quantamagazine.bsky.social article on our theory of creativity in convolutional diffusion models lead by @masonkamb.bsky.social See also our paper with new results in version 2: arxiv.org/abs/2412.20292 to be presented as an oral at @icmlconf.bsky.social #icml25

2nd Workshop on Test-Time Adaptation: Putting Updates to the Test! (PUT) Our second workshop on test-time adaptation will be held at ICML 2025 this summer in Vancouver!

The accepted papers are out, but we're not done yet! 💥

Introducing our "Bonus Track": Got updates, teasers, or fresh insights? Submit a 1-slide spotlight of your work by July 14 and get featured during the workshop!

Find more here: tta-icml2025.github.io

@icmlconf.bsky.social

#icml25

So kudos to 🧑‍🚀 Filippo and @marceldotsci.bsky.social bsky.app/profile/labc..., thanks to funders @erc.europa.eu @snf-fns.ch @cscsch.bsky.social and @nccr-marvel.bsky.social, and go to Filippo's talk if you are at #icml25, to see this nice mix of #compchem and #machinelearning!

Fujimoto et al - Allo / auto in ENKT? Japanese data.
• if in CR, ASCT appear superior to allo
• if in PR, allo appears preferable
• no impact on source of allo cells
I normally allo these pts in CR1 so potentially practise changing.
#18ICML #ICML25 #lymsm

Held et al - NIVEAU trial. Nivo-GemOx vs GemOx RR PTCL
• 76 pts part of bigger trial
• PFS no diff but was ⬆️ when MVA performed
• responses ⬆️
• BUT ⬆️ toxic incl ‘infn’ (?hyperinfl state instead); ⬆️ deaths.
Doubt there will be further development.
#18ICML #ICML25 #lymsm

Patel: JNJ4496 (anti CD19/20 tandem CAR) RR DLBCL, 41BB costim
• global Ph 1B
• med age 72y
• at RP2D, ORR 100%, CRR 80%
• no G3+ CRS, 8% ICANs
Very impressive results
#18ICML #ICML25 #lymsm

Dreyfuss et al
• ISRT & Pembro in localised relapsed cHL
• RT dose varied according to response to Pembro
• generally safe & good outcomes
• did not seem to impair subseq Rx
Consider for older relapsed pts
#18ICML #ICML25 #lymsm

Dreyfuss et al
• ISRT & Pembro in localised relapsed cHL
• RT dose varied according to response to Pembro
• generally safe & good outcomes
• did not seem to impair subseq Rx
Consider for older relapsed pts
#18ICML #ICML25 #lymsm

Davison et al - PRO substudy of S1826 substudy. Interesting outputs:
• NivoAVD showed more fatigued compared with BV-AVD on treatment but did not meet ‘clin significant’ threshold.
• BV-AVD showed excess neuropathy which had a clinically meaningful impact.
#ICML25 #lymsm

Evans et al: E-HIPI prognostic score early cHL
• Holistic consortium
• 1 binary factor: female sex better prog
• 3 continuous: max tumour diam, Hb and albumin. Great work from

Online calculator now available!
#18ICML #ICML25 #lyms

Great work by @ABarrettHaem. NHSE data, older #Hodgkin pts. Outcomes ⬆️ with dox. Clearly may reflect selection, but interesting assocn seen between regional difference in dox use and outcome.

Take home message: if safe to use dox then do use it!

#18ICML #ICML25 #lymsm

Lewis et al - MRD analysis in ENRICH: IR vs R-chemo in 1L MCL
• flow cytometry based
• MRD- most rapid with BR
• with RCHOP, even MRD- pts do poorly
• with IR, no assocn with MRD & outcome
#18ICML #ICML25 #lymsm

Flerlage et al
• GLOW data nod LPHL
• completely resected 1A, RT assoc with ⬆️ PFS compared with observation
• unresected - ABVD-RT & RT do v well
• no infield relapses if RT given
Great work which significantly informs practice. #18ICML #ICML25
#lymsm

Rossi et al - baseline MTV, Dmax & esp combo, assoc with PFS.
Independent effect of IPSS.
Both factors improve prognostic impact of iPET.
Based on AHL2011 study so all have 2x eacBEACOPDac as 1st 2 cycle.
#18ICML #ICML25

Herrera et al - ctDNA in S1826
• 97% ctDNA + baseline
• clearance at C3D1 assoc with PFS
• within pos pts, those with minor ⬇️ assoc with esp poor outcomes
• EOT ctDNA detection also assoc with poor outcomes
CtDNA ripe for response adapted trials.
#18ICML #ICML25 #lymsm

Ahmed et al - EBV & cHL subtype in S1826 study
• EBV+ pts do much better with NAVD compared with BV-AVD (EBV- also do better but less marked)
• non-NS histology also benefited more from NAVD.
Biology is important!
#lymsm #18ICML #ICML25

Things are hotting up for the first of 2 #Hodgkin lymphoma sessions today.
#18ICML #ICML25 #lymsm

🧵 ✨ Hide & Seek: Transformer Symmetries Obscure Sharpness & Riemannian Geometry Finds It ✨

Excited to announce our paper on factoring out param symmetries to better predict generalization in transformers ( #ICML25 spotlight! 🎉)

Amazing work by @marvinfsilva.bsky.social and Felix Dangel.
👇

AI Scaling: From Up to Down and Out AI Scaling has traditionally been synonymous with Scaling Up, which builds larger and more powerful models. However, the growing demand for efficiency, adaptability, and collaboration across diverse a...

🎉 Yunke Wang, Yanxi Li and Chang Xu's paper "AI Scaling: From Up to Down and Out" has been accepted at #ICML25! Check it out here: arxiv.org/abs/2502.01677

🔗 More about ICML 2025: icml.cc
#Research #AI #MachineLearning #ScalingLaws

Interested in Data-centric AI?? Check out our new @icmlconf.bsky.social workshop 🥳🥳

#icml25