Sign in

Janis Keuper 🇮🇱

@janiskeuper.bsky.social
48 followers 79 following 26 posts

Research in Machine Learning, Computer Vision and Geophysics. www.keuper-labs.org

PostsRepliesMedia
Reposted by Janis Keuper 🇮🇱
Henry Mance @henrymance.ft.com · 19/08/2026
Woah.
18362691865
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 21/08/2026
Congratulations to our PhD student Pius for the ICPR 2026 "Best Scientific Paper Award". 🥳 Pius is working on the the critical document parsing step in LLM/VLM pipelines. Paper: link.springer.com/content/pdf/...
001
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 08/07/2026
We present two papers on Thursday at @icmlconf.bsky.social : Poster session 6: Poster #3107 "We need to re-think the concept of 'real' images" icml.cc/virtual/2026... Poster session 7: Poster #1114 "mSOP-765k: A Benchmark For Multi-Modal Structured Output Predictions" icml.cc/virtual/2026...
icml.cc
ICML Poster Position: We need to re-think the concept of “real” images.ICML 2026
211
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 20/05/2026
New TMLR paper out: "A New Kind of Network? Review and Reference Implementation of Neural Cellular Automata." openreview.net/pdf?id=NRwjj...
111
Reposted by Janis Keuper 🇮🇱
Markus Eichhorn @markuseichhorn.bsky.social · 07/05/2026
Bingo! Do I win a prize now?
Comic strip titles self-sabotaging academic habits that feel normal, featuring nine panels.
1122583
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 04/05/2026
Our #ICML26 position paper "We need to re-think the concept of “real” images." is now on Arxiv: arxiv.org/pdf/2509.21864
101
Reposted by Janis Keuper 🇮🇱
DAGM GCPR 2026 @gcpr-by-dagm.bsky.social · 16/03/2026
We are excited to share our call for papers with a submission deadline on May 21st, 2026! We invite submissions of high-quality research papers presenting original contributions in all areas of pattern recognition! Read more: www.gcpr-vmv.de/year/2026/gc... #GCPR2026 #VMV2026
gcpr-vmv.de
GCPR VMVGCPR VMV (Call for Papers)
174
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 18/02/2026
We are proud to be part of the @ellis.eu community. @margretkeuper.bsky.social has been an Ellis Fellow for some time, now @janiskeuper.bsky.social also joined as a Member...
001
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 21/01/2026
Open PostDoc position: Hardware aware NAS / Model Distillation / Model Compression of Vision and Language Models for Humanoid Robots... www.keuper-labs.org/team/#open-p...
021
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 17/01/2026
New paper out. (Accepted at TMLR with J2C Certification!): mSOP-765k: A Benchmark For Multi-Modal Structured Output Predictions. www.msop-765k.org
011
Reposted by Janis Keuper 🇮🇱
Andrew Heiss @andrew.heiss.phd · 15/12/2025
Grading and googling hallucinated citations, as one does nowadays, and now that LLMs have been around for a while, I've discovered new horrors: hallucinated journals are now appearing in Google Scholar with dozens of citations bc so many people are citing these fake things
13039551264
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 15/11/2025
A small step for science, quite a big one for us: reaching 10k citations. Thanks to all co-authors! scholar.google.de/citations?hl...
011
Reposted by Janis Keuper 🇮🇱
angela zhou @angelamczhou.bsky.social · 31/10/2025
blog.arxiv.org/2025/10/31/a... FYI the blog post for the updated policy is out. Our llm future is dire:/
blog.arxiv.org
Attention Authors: Updated Practice for Review Articles and Position Papers in arXiv CS Category – arXiv blog
3276
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 09/10/2025
Looks like I belong to the big boys now 🫡 Mail from OpenAI: "Congratulations on processing over 10 billion tokens with the OpenAI API." 😅 A lot of these tokens went into this project: "Prompt Injection Attacks on LLM Generated Reviews of Scientific Publications" arxiv.org/pdf/2509.10248
arxiv.org
021
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 09/10/2025
Our new paper (accepted at #NeurIPS UrbanAI Workshop): "Real-time Prediction of Urban Sound Propagation with Conditioned Normalizing Flows" arxiv.org/pdf/2510.04510 is part of our work using generative model to predict complex physics www.physics-gen.org
031
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 02/10/2025
Our GCPR @gcpr-by-dagm.bsky.social paper (Oral) "Assessing Foundation Models for Mold Colony Detection with Limited Training Data" is now on Arxiv: arxiv.org/pdf/2510.00561
022
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 02/10/2025
Our #NeurIPS ORAL "MaxSup: Overcoming Representation Collapse in Label Smoothing", joint work with @cispa.de, is now on Arxiv: arxiv.org/pdf/2502.15798
021
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 16/09/2025
Following up on the discussions about prompt injections which were found in Arxiv paper sources - indicating that some authors try to manipulate LLM generate reviews: beyond the ethical implications, I was wondering if this actually works and did some experiments... arxiv.org/abs/2509.10248
arxiv.org
Prompt Injection Attacks on LLM Generated Reviews of Scientific Publications
The ongoing intense discussion on rising LLM usage in the scientific peer-review process has recently been mingled by reports of authors using hidden prompt injections to manipulate review scores. Sin...
101
Reposted by Janis Keuper 🇮🇱
Kendra Albert @kendraserra.bsky.social · 11/09/2025
Very glad to see that someone is doing the important work of aligning AI alignment. alignmentalignment.ai
alignmentalignment.ai
Center for the Alignment of AI Alignment Centers
We align the aligners
22412106
Reposted by Janis Keuper 🇮🇱
Dare Obasanjo @carnage4life.bsky.social · 20/08/2025
MIT’s NANDA initiative found that 95% of generative AI deployments fail after interviewing 150 execs, surveying 350 workers, and analyzing 300 projects. The real “productivity gains” seem to come from layoffs and squeezing more work from fewer people not AI.
fortune.com
MIT report: 95% of generative AI pilots at companies are failing
There’s a stark difference in success rates between companies that purchase AI tools from vendors and those that build them internally.
7746041877
Reposted by Janis Keuper 🇮🇱
Christian Wolf @chriswolfvision.bsky.social · 23/07/2025
Update on hidden prompts in papers targeting LLM reviews: ICML 2025 PCs react. icml.cc/Conferences/...
44213
Reposted by Janis Keuper 🇮🇱
Randall Munroe @xkcd.com · 22/07/2025
Replication Crisis xkcd.com/3117/
4-panel comic. (1) [Person 1 with ponytail flanked by person with short hair and another person speaking into microphone at podium] PERSON 1: In the early 2010s, researchers found that many major scientific results couldn’t be reproduced. (2) PERSON 1: Over a decade into the replication crisis, we wanted to see if today’s studies have become more robust. (3) PERSON 1: Unfortunately, our replication analysis has found exactly the same problems that those 2010s researchers did. (4) [newspaper with image of speakers from previous panels] Headline: Replication Crisis Solved
284843654
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 22/07/2025
🚀 Calling all ML, CV and Sensor Enthusiasts! Join us at the L2S: Learning to Sense Workshop at #NeurIPS2025 ! 🗓 Date: December 6/7, 2025 📍 Location: San Diego, USA sites.google.com/view/l2s-wor...
032
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 15/07/2025
Attending #ICML2025? Watch out for our paper "DCBM: Data-Efficient Visual Concept Bottleneck Models" presented by @katharinaprasse.bsky.social -> github.com/KathPra/DCBM
031
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 08/07/2025
Congratulations to @paulgavrikov.bsky.social for an excellent PhD defense today!
032
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 30/06/2025
Two papers with @keuper-labs.bsky.social participation accepted at #ICCV2025: 1) Scientific figure generation with TikZero, it generates scientific figures from text as high-level, human-interpretable, and editable graphics programs -> arxiv.org/pdf/2503.11509
131
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 23/06/2025
Searching for latex symbols? I found this handy tool: detexify.kirelabs.org/classify.html
000
Reposted by Janis Keuper 🇮🇱
DAGM GCPR 2026 @gcpr-by-dagm.bsky.social · 21/06/2025
🚨 Calling all Vision and ML researchers! 🚨 Missed the regular GCPR deadline? No worries, the Fast Review Track deadline is July 1, 2025 (11:59 PM CEST). Polish up that revision from a previous submission and submit now! 🔄✍️ #GCPR2025 #ComputerVision #AIResearch
033
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 14/06/2025
Presenting our poster #45 today at #CVPR in poster session 3: "PhysicsGen: Can Generative Models Learn from Images to Predict Complex Physical Relations?" cvpr.thecvf.com/virtual/2025...
111
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 11/06/2025
A Visual RAG Pipeline for Few-Shot Fine-Grained Product Classification @bianca-lamm.bsky.social @janiskeuper.bsky.social Poster Wed. 2-4pm at ExHall D Fine-Grained Visual Categorization Workshop at #CVPR2025 ( sites.google.com/view/fgvc12) arxiv.org/pdf/2504.11838?
122
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 11/06/2025
Main #CVPR2025 Conference Paper: PhysicsGen - Can Generative Models Learn from Images to Predict Complex Physical Relations? Martin Spitznagel, Jan Vailant, @janiskeuper.bsky.social Sat 14 Jun 10:30 a.m. CDT — 12:30 p.m. ExHall D Poster #45 cvpr.thecvf.com/virtual/2025...
022
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 11/06/2025
We are presenting 7 papers (main conference+ workshops) at #CVPR2025 www.keuper-labs.org/research/?se...
563
Reposted by Janis Keuper 🇮🇱
DAGM GCPR 2026 @gcpr-by-dagm.bsky.social · 08/05/2025
🚀 Exciting news! #GCPR2025 – the DAGM German Conference on Pattern Recognition – is coming to Freiburg, Germany! 📍🗓️ Mark your calendars: September 23-26, 2025!
1168
Reposted by Janis Keuper 🇮🇱
François Fleuret @francois.fleuret.org · 28/04/2025
I asked "on the other platform" what were the most important improvements to the original 2017 transformer. That was quite popular and here is a synthesis of the responses:
420643
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 22/04/2025
Visit our poster presentations at #ICLR2025 iclr.cc/virtual/2025...
132
Reposted by Janis Keuper 🇮🇱
Paul Gavrikov @paulgavrikov.bsky.social · 21/04/2025
On Thursday, I'll be presenting our paper "Can We Talk Models Into Seeing the World Differently?" (#328) at ICLR 2025 in Singapore! If you're attending or just around Singapore, l'd love to connect—feel free to reach out! Also, I’m exploring postdoc or industry opportunities–happy to chat!
145
Reposted by Janis Keuper 🇮🇱
Julia Grabinski @juliagrabinski.bsky.social · 21/04/2025
Ever wondered how big convolutional filters should really be? 🤔 We push things to the limit literally. Infinite-size convs, but efficient. “As Large as It Gets – Studying Infinitely Large Convolutions via Neural Implicit Frequency Filters” at #ICLR2025, Singapore! #ML #DeepLearning #ComputerVision
273
Reposted by Janis Keuper 🇮🇱
NeurIPS Blog – NeurIPS conference blog [Unofficial] @blog.neurips.cc.web.brid.gy · 02/04/2025
blog.neurips.cc
NeurIPS 2025 Position Paper Track
Track Chairs: Barbara Engelhardt, Cynthia Rudin This year at NeurIPS, we are initiating a Position Paper Track, following ICML 2024 and 2025 Position Paper Tracks. This is a time of accelerated development and deployment of methods designed in the NeurIPS community, where many people interact with AI in myriad ways – including high-stakes settings. Our […]
021
Reposted by Janis Keuper 🇮🇱
Sung Kim @sungkim.bsky.social · 31/03/2025
You’ve probably heard about how AI/LLMs can solve Math Olympiad problems ( deepmind.google/discover/blo... ). So naturally, some people put it to the test — hours after the 2025 US Math Olympiad problems were released. The result: They all sucked!
917350
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 01/04/2025
New pre-print: "FOUNDATION MODELS FOR SEISMIC DATA PROCESSING: AN EXTENSIVE REVIEW". Interesting findings for #Seismic / #Geophysic Applications: I) hierarchical models outperform basic approaches, II) pre-training on natural images actually helps in low-data settings. arxiv.org/pdf/2503.02113
001
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 14/03/2025
@hf.co featured our latest CVPR paper on the physical correctness of generative model in daily papers: huggingface.co/papers/2503.... and is hosting the benchmark data: huggingface.co/datasets/msp... Project page: www.physics-gen.org
101
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 10/03/2025
Our latest paper accepted at #CVPR2025 is now arxiv: "PhysicsGen: Can Generative Models Learn from Images to Predict Complex Physical Relations?" arxiv.org/abs/2503.05333
111
Reposted by Janis Keuper 🇮🇱
Yoshua Bengio @yoshuabengio.bsky.social · 29/01/2025
Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU. It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵 Full Report: assets.publishing.service.gov.uk/media/679a0c... 1/21
7256104
Reposted by Janis Keuper 🇮🇱
Paul Gavrikov @paulgavrikov.bsky.social · 27/01/2025
Proud to announce that our paper "Can We Talk Models Into Seeing the World Differently?" was accepted at #ICLR2025 🇸🇬. This marks my last PhD paper, and we are honored that all 4 reviewers recommended acceptance, placing us in the top 6% of all submissions.
134
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 25/01/2025
We will be presenting 3 papers at @iclr-conf.bsky.social : 1. Can we talk models into seeing the world differently? 2. VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis 3. "As large as it gets – Studying Infinitely Large Convolutions via Neural Implicit Frequency Filters"
134
Reposted by Janis Keuper 🇮🇱
Chris Offner @chrisoffner3d.bsky.social · 07/01/2025
With 128 GB of unified memory, NVIDIA's new Project DIGITS finally offers a desktop machine that can compete with my MacBook Pro on memory, but outcompetes it on the "lack of CUDA support" front. nvidianews.nvidia.com/news/nvidia-...
2212
Reposted by Janis Keuper 🇮🇱
Chris Offner @chrisoffner3d.bsky.social · 10/12/2024
"Sora is a data-driven physics engine." x.com/chrisoffner3...
1213716
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 26/11/2024
Congratulations to Mario Fernandez for his successful defense in Paris at ENS - "Towards deep learning for seismic processing". Joint supervision with Matthias Delescluse (ENS) FraunhoferITWM and @keuper-labs.bsky.social
001
Janis Keuper 🇮🇱 @janiskeuper.bsky.social · 23/11/2024
New IJCV paper out: Reliable Evaluation of Attribution Maps in CNNs: A Perturbation-Based Approach. Joint work with Fraunhofer ITWM and @keuper-labs.bsky.social link.springer.com/content/pdf/...
011
Reposted by Janis Keuper 🇮🇱
Keuper Labs @keuper-labs.bsky.social · 23/11/2024
www.keuper-labs.org is going blue ... we will post our latest research and other #ComputerVision and #MachineLearning news in our new #BlueSky channel ...
022