📢 Our recent paper just got accepted at #ICML2026!
𝗧𝗶𝘁𝗹𝗲: The Expert Strikes Back: Interpreting Mixture-of-Experts Language Models at Expert Level
𝗟𝗶𝗻𝗸: https://t.co/7GIGocKMXq
More info 👇
#LLM#MoE#MechanisticInterpretability
1/ Excited to share that our paper “The Expert Strikes Back” has been accepted to #ICML2026!
We ask: does sparse routing in MoE language models make them easier to interpret?
Paper: https://t.co/1AbTcIsnn1
📢 Our recent paper and collaboration just got published at Neural Computing and Applications
𝐓𝐢𝐭𝐥𝐞: A parameter-free adaptive resonance theory-based topological clustering algorithm capable of continual learning
𝐋𝐢𝐧𝐤: https://t.co/G5VWToH1Tl
#ContinualLearning#AI#ML
Our paper has been accepted to Neural Computing and Applications.
We propose a parameter-free ART-based topological clustering algorithm for continual learning.
The method automatically estimates key thresholds, eliminating manual tuning.
Paper: https://t.co/lneyhRyKSs
Our paper "Agentic Skill Discovery" is now published in Robotics and Autonomous Systems. A method for #robots to learn new skills from scratch using #LLM reasoning, #RL and #VLM validation.
Link to paper: https://t.co/GI5yLWDcbe
#Robotics#AI#MachineLearning#SkillDiscovery
The Knowledge Technology Team contributed very well to the 34th International Conference on Artificial Neural Networks (#ICANN2025) with various papers and talks!
🔗https://t.co/s4t50Tibot
🔗https://t.co/FPtG0e3mwE
🔗https://t.co/0YxfzPgN7G
#Robotics#HRI#NeuralNetworks#ML
Here's our paper "Key-Point-based Diffusion for Robotic Motion Planning" from #ICANN2025
- Collision-free motion plans from traditional planners
- Faster motion planning
- Diffusion-based no-collision plan generation
- Trained and evaluated on #NICOL
🔗 https://t.co/FPtG0e2OH6
Check out our paper "Pointing-Guided Target Estimation via Transformer-Based Attention" from #ICANN2025
- Teaching the #robot pointing gestures for #HRI
- Transformer model for pose-object reasoning
- Evaluated with #NICOL in a collaboration scenario
🔗https://t.co/s4t50Tibot
📢#ICANN2025 has kicked off in Kaunas, Lithuania!
ENNS President Prof. Stefan Wermter opened the 34th International Conference on Artificial Neural Networks, the flagship event of the European Neural Network Society.
🔗More: https://t.co/cPnlX1MSFW
#AI#NeuralNetworks#ENNS
📢see here collaboration between @ComeniusUni & @RBCS_CONTACT on novel neural network-based method creating precise, smooth, and customizable robotic arm trajectories without requiring large datasets applied to NICO @KnowTechUHH enabling accurate pointing and gesture movements⬇️
📢Our paper "Shaken, Not Stirred" just got accepted at #IROS2025
🍸 Novel real-world dataset: glass object detection in #robotics
⚙️ Auto-labeling pipeline: rapid data annotation
🤖 Real-time integration with #NICOL: a human-robot bartending scenario
🔗 https://t.co/0YxfzPgN7G
📢 #ICANN25 Special Session on #Neurorobotics
🤖 Interdisciplinary research in ML, robotics and bio-inspired AI
🧠 Topics: Neural control, cognitive architectures, human-robot interaction, and more
📅 Paper submission deadline: April 30, 2025
🔗More info: https://t.co/UyQ2Mf5d29
Our work presented at #ICSR24
💬#LLM to predict human intent via speech & gestures
🧠Multimodal cues for real-time reasoning
🤖Object categorization task with the #NICOL robot
📰Now published with #OpenAccess
🔗https://t.co/ijak8jgZm7
@TERAIS_project
Our newest paper and model NOVIC presented at WACV'25
💬 Classifies any image via noun generation
🏞️ Produces diverse high-fidelity classifications without any input prompts
🚀 Trained on text only, zero-shot transfer to images
🏎️ Extremely fast and efficient
#WACV2025#AI#ML
FabuLight-ASD is an advanced active speaker detection model integrating face, audio & body pose for better accuracy & robustness, outperforming Light-ASD.
Paper (Open access): https://t.co/UnFGUFqbKG
Code: https://t.co/pp7Gl7NkCt
New paper📰
Our new paper 𝐹𝑎𝑏𝑢𝐿𝑖𝑔ℎ𝑡-𝐴𝑆𝐷: 𝑢𝑛𝑣𝑒𝑖𝑙𝑖𝑛𝑔 𝑠𝑝𝑒𝑒𝑐ℎ 𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝑣𝑖𝑎 𝑏𝑜𝑑𝑦 𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒 just got published in Neural Computing and Applications Journal by Springer
Links👇
#ASD#LightASD#ActiveSpeakerDetection
34th International Conference on Artificial Neural Networks #ICANN2025
Call for Papers, Special Sessions, Workshops & Tutorials
Kaunas, Lithuania
Sep 9 - 12, 2025
See: https://t.co/pjqyBoezQb
#ENNS#AI#NeuralNetworks#ML#MachineLearning