Dawid

@ddjrym

Joined January 2021

265 Following

83 Followers

Replies Reposts Media

Dawid
@ddjrym
Oct 15
Have a look! That is truly amazing part of the work we did recently. And the results exceeded our expectations! Great job to all involved ;)
Siddhartha Gairola
@sidgairo18
Oct 15
🚨 Checkout our new work on learning self-supervised visual part representations: "DisParQ: Self-Supervised Part Concepts for Interpretable Vision Foundation Models". Very proud and happy with this work. Tremendous effort by all involved! 🤗 🚀 1/10 📄 arxiv.org/abs/2610.09802
Dawid
@ddjrym
Sep 25
🚀 Excited to share that B-XAIC has been accepted at NeurIPS 2026! Current XAI benchmarks for GNNs in chemistry are flawed. B-XAIC fixes this: 🔹 50K real molecules (ChEMBL) 🔹 7 tasks w/ ground-truth explanations 🔹 Rigorous XAI eval 📦See: github.com/mproszewska/B-… #NeurIPS2026
GitHub - mproszewska/B-XAIC
From github.com
1
Dawid
@ddjrym
Jul 2
#ICML2026 is coming! Will be presenting the poster on Tuesday at 2pm! Please stop by or reach out to discuss the work or how science can look like in the era of Agents 😏
Siddhartha Gairola
@sidgairo18
Jul 1
✈️Will be attending #ICML2026 in Seoul, South Korea 🇰🇷 to present our Spotlight project, DAVE 🏅 We fix a persistent issue in ViT explainability: ❌unstable, artifact-heavy pixel attributions. DAVE yields: ✅fine-grained pixel-level maps ✅without patch-grid saliency. Come
1
Dawid
@ddjrym
Apr 30
That is a very nice surprise. ICML spotlight! Something that was not even in my dreams when doing a PhD. Adam, @sidgairo18, @BartosZielinski and others! You are the one behind this success!
Siddhartha Gairola
@sidgairo18
Apr 30
DAVE gets into ICML 2026 as a spotlight! 💥🚀 Very heartening to share ... code and updated manuscript to be released very soon! cc: @ddjrym Adam et al.
1
Dawid
@ddjrym
Feb 10
Our recent collab! Great teamwork with amazing results for attribution methods with ViTs! Check this out :)
Siddhartha Gairola
@sidgairo18
Feb 10
🚀New preprint: DAVE — Distribution-aware Attribution via ViT Gradient DEcomposition. 1/11 🔍 What’s new: We fix a persistent issue in ViT explainability: unstable, artifact-heavy pixel attributions. DAVE yields fine-grained pixel-level maps without patch-grid saliency.

Dawid

Dawid’s posts