Evangelos Kazakos
@ekazakos.bsky.social
Postdoctoral researcher @ CIIRC, CTU, Prague working in vision & language. Also robotics noob. PhD from University of Bristol. Ex. Samsung Research (SAIC-C). I love coffee and plants. And socks. 𖤓꩜𓆉𓃰𖤐 https://ekazakos.github.io/
It is strange how much LLMs turned out to be able to solve such a wide range of hard problems that would not, instinctively, seem to be problems that a model of human language would be able to solve This is from a Stanford project that let Astra drive a robot in a kitchen tml.stanford.edu/homebody/
How about if they visit Gaza? Oh but you guys don't let people do that do you?
The Irish Times@irishtimes.com · 3d agoIsrael captain Dor Peretz says Irish players should visit Israel then ‘can judge and think’
AI for Peace 2nd edition is back at @neuripsconf.bsky.social, Paris, December. The call for in person presentations is open. Submit, come, and be part of a community that is building up. AI for Peace @ NeurIPS 2026, Paris - One-page, non-archival, talk or poster. - Deadline 14 September 2026
Does it sound a bit like Jev is probably in the direction of combining Transformers/LLMs with neurosymbolic reasoning or is it only me?
If I found login credentials for a US government website and accessed it, I would go to jail. Just saying
Excited to share that our paper ZendoWorld has been accepted to #NeurIPS2026! 🎉 ZendoWorld tests interactive visual reasoning and demonstrates that current agents (VLM and NeSy) still struggle with both rule induction and informative experimentation 🤖💭
Antonia Wüst@toniwuest.bsky.social · 3mo agoCan AI agents act like scientists when reasoning about visual scenes? 🔬 Not just label what they see, but form explicit hypotheses, design experiments to test them, and revise their beliefs in light of the evidence. We built ZendoWorld to find out!
The PyTorch Foundation OSPO & Academic Outreach Working Group is seeking academic open source projects for a Day 0 workshop at PyTorchCon NA 2026. Submit by Sept. 30: https://pytorch.org/blog/from-research-project-to-open-source-ecosystem-bring-your-academic-pytorch-project-to-pytorchcon-na/
1/ Matryoshka Attribution uses gradient descent plus causal interventions to localize the parts of a network responsible for a behavior; the authors claim a large MIB jump. X: https://x.com/aryaman2020/status/2102800933659000846 Paper: https://arxiv.org/abs/2609.25518
Quick belated paper promotion. We had 2 papers at ECCV that I was in Malmo to present! UniFlow is a foundation model for 3D LiDAR scene flow, demonstrating zero-shot generalization across datasets and LiDAR sensors for the first time. We also showed that semantics don’t generalize the same way.
The appropriate measure when somebody asks for too many CPUs on a single GPU job and the other GPUs of that node are now idling.
During my Euro sabbatical, I’ve been investing into impacting local AI policy & media. In this @deutschlandfunk.de.web.brid.gy show I (in stress-activated German!) debunk alarmist AI doom narratives. Actionable priority problems rn: #oligarchy, #marketconcentration, #regcapture, antidem #ideology.
🏹 Job alert: PhD Position in Physical AI at Aarhus University 📍 Aarhus 🇩🇰 ⏰ 1 Nov 🔗 https://bit.ly/3ULRPwx
PhD Position in Physical AI: Adaptive Foundation Models for Robotics
bit.ly
A lesson from decades of work on automation is that autonomous systems require continual monitoring and human oversight. System designers never(*) succeed in anticipating all of the ways that a system can fail.
The conclusion is that we must design our "autonomous" systems to be combined human-machine systems with continual supervision from the very beginning.
I have been yelling since 2021 about how using synthetic text extruding machines for information access is a TERRIBLE IDEA. And yet and yet and yet: they are marketing so heavily that way that people keep falling for it. edition.cnn.com/2026/09/18/p...
Exclusive: US military had close call after using AI for false intelligence report, sources say | CNN Politics
The episode shows the risks of using this new, relatively poorly understood technology in the middle of the Iran war
edition.cnn.com
I am SORRY for sharing something that was posted on the other place, but this is too good.
🏹 Job alert: Fully Funded PhD Positions in Machine Learning & AI for Science 📍 Vienna 🇦🇹 ⏰ 1 Nov 🔗 https://bit.ly/4ipJUyD
The AITHYRA International PhD Call in AI for Biomedicine (all genders)
Do you want to help transform human health through machine learning and life science approaches? Join AITHYRA in Vienna for a fully funded PhD at the intersection of machine learning, experimental bio...
aithyra.at
𝗧𝗿𝗮𝗰𝗸, 𝗔𝗿𝘁𝗶𝗰𝘂𝗹𝗮𝘁𝗲, 𝗔𝗰𝘁: 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗔𝗿𝘁𝗶𝗰𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝗳𝗿𝗼𝗺 𝗖𝗮𝘀𝘂𝗮𝗹 𝗛𝘂𝗺𝗮𝗻 𝗩𝗶𝗱𝗲𝗼𝘀 Jiaming Zhang, Homanga Bharadhwaj arxiv.org/abs/2609.19119 Trending on scholar-inbox.com
The response in Israel to Naza is beyond unhinged. Mobs have shown up at the house of co-director Rachel Szor's family. All over social media the directors are being branded traitors and "not Jews". There is a frenzy of incitement to violence against them online www.theguardian.com/world/2026/s...
Netanyahu seeks to strip citizenship from Israelis as threats to NAZA film directors intensify
Israeli prime minister to pursue new laws as directors of Guardian-produced film face threats from far-right officials and activists
theguardian.com
How can vision models adapt to different compute budgets? ⚙️ In this ELLIOT Research Spotlight 🔬, we highlight SnapViT, a method from the University of Amsterdam, University of Technology Nuremberg & NVIDIA that makes pretrained Vision Transformers elastic without retraining. Paper in comments 👇
Federal health officials are pushing to reimburse autonomous clinical A.I. via Medicare, but clinical validation isn't keeping up with the rush. The FDA has cleared 1,500+ narrow algorithms, but none of these prescribing agent pilots have formal agency approval yet. #MedSky
U.S. Health Officials Move Quickly to Deploy Medical A.I. Despite Concerns (Gift Article)
Federal projects are rolling out A.I. agents that offer therapy and prescribe medicine, raising concerns about safety and the influence of venture capital investors.
nytimes.com
It was a great week @eccv.bsky.social #ECCV2026 #Malmo for current members @compscibristol.bsky.social #MaVi group and catching up with alumni! Here's a group pic missing a few, @prajwalgatti.bsky.social @zhifan-zhu.bsky.social @ekazakos.bsky.social Siddhant, Ahmad, Jacob, Alex, Toby, Kevin.
First slide: “AI sucks!” I will start to like the guy in the end 😂
