
Explainable Machine Learning
@ExplainableML
Institute for Explainable Machine Learning @HelmholtzMunich and Interpretable and Reliable Machine Learning group @TU_Muenchen
내가 좋아할 만한 콘텐츠
2 papers accepted at NeurIPS 2025 🎉 🔹 Manipulating Feature Visualizations with Gradient Slingshots 🔹 Capturing Polysemanticity with PRISM: A Multi-Concept Feature Description Framework

Reward hacking is challenging when fine-tuning few-step Diffusion models. Direct fine-tuning on rewards can create artifacts that game metrics while degrading visual quality. We propose Noise Hypernetworks as a theoretically grounded solution, inspired by test-time optimization.
💫 After four PhD years on all things multimodal, pre- and post-training, I’m super excited for a new research chapter @GoogleDeepMind 🇨🇭! Biggest thanks to @zeynepakata and @OriolVinyalsML for all the guidance, support, and incredibly eventful and defining research years ♥️!



United States 트렌드
- 1. Rickey 2,325 posts
- 2. Waddle 3,455 posts
- 3. Big Balls 22.8K posts
- 4. Argentina 498K posts
- 5. #BeyondTheGates 5,528 posts
- 6. $HIMS 3,961 posts
- 7. Kings 157K posts
- 8. Olave 2,986 posts
- 9. Russell Westbrook 13.6K posts
- 10. Voting Rights Act 26.1K posts
- 11. Maybe in California N/A
- 12. Justice Jackson 18.5K posts
- 13. #ClockTower1Year N/A
- 14. #TrumpsShutdownDragsOn 5,997 posts
- 15. Hayley 5,198 posts
- 16. Aphrodite 4,444 posts
- 17. Veo 3.1 5,792 posts
- 18. Meyers 2,366 posts
- 19. annabeth 2,387 posts
- 20. Capitol Police 26.5K posts
내가 좋아할 만한 콘텐츠
-
Zeynep Akata
@zeynepakata -
ELLIS
@ELLISforEurope -
AI Conference DL Countdown
@DlCountdown -
Mihaela van der Schaar
@MihaelaVDS -
Michael Black
@Michael_J_Black -
Been Kim
@_beenkim -
Cambridge MLG
@CambridgeMLG -
Yarin
@yaringal -
UvA AMLab
@AmlabUva -
Oana-Maria Camburu
@oanacamb -
Accepted papers at TMLR
@TmlrPub -
Georgios Pavlakos
@geopavlakos -
Federico Tombari
@fedassa -
Christoph Molnar 🦋 christophmolnar.bsky.social
@ChristophMolnar -
Kate Saenko
@kate_saenko_
Something went wrong.
Something went wrong.