ALT We hosted a talk by Sam Carton on using #AgenticAI for scientific discovery. From designing high-entropy alloys to tackling multi-objective optimization, he explored how #LLMs & #AI agents can address complex #materialdesign challenges. Watch now: https://youtu.be/hnMP-utXUA8. #NECLabs

ALT Our latest work shows that simple truncation can often match & sometimes outperform MRL for compressing text embeddings. Across 24 retrieval & classification tasks, MRL only showed clear benefits under extreme compression. Learn more: https://neclab.eu/research-groups/human-centric-ai/publications#c3537 #NECLabs #AI #EMNLP

ALT Our self-adaptive memory control framework helps #AI agents manage memory more efficiently through uncertainty-guided retrieval & decay-driven activation, improving long-term reasoning while reducing costs. Learn more in our #EMNLP-accepted paper: https://neclab.eu/research-groups/human-centric-ai/publications#c1348 #NECLabs

ALT We hosted a talk by @gneubig on whether #AI can reliably assess scientific research. Study of 45 scientists shows AI flags issues in methods & reproducibility but misses field norms, raising questions on impact-focused research evaluation. Watch: https://youtu.be/-l2ufFjW5BA?si=nxhpjvXNqQH5m0kl. #NECLabs

ALT We hosted a talk by @egere14 and C. Greisinger from @utn_nuremberg on #multimodal scientific content generation. They discussed using #AI to generate & edit scientific figures (e.g. via TikZ or Python code) from captions, text & sketches. Watch here: https://youtu.be/1X1thAlpYsE. #NECLabs

ALT Our work shows that #AI reviewers can rival human performance on key review tasks but are most effective when combined with human expertise. Learn more: https://neclab.eu/research-groups/human-centric-ai/publications#c2760. #NECLabs #PeerReview

ALT Our recent work shows that using reasoning models as evaluators improves evaluation quality with additional test-time compute, enabling stronger re-ranking of #languagemodel outputs & matching the gains of increased compute at generation time. Learn more: https://neclab.eu/research-groups/human-centric-ai/publications#publications-2202 #NECLabs

ALT We hosted an insightful talk by Dr. Zhiyong Lu on how evidence-grounding strategies can improve the reliability, interpretability and scientific soundness of #LLMs for #biomedical research and #clinicalapplications. Watch here: https://youtu.be/GmE8uAV_S7M. #NECLabs #AI

ALT Our latest work introduces a framework for exact logical guidance in #diffusionmodels. It enables principled composition of complex #Boolean constraints at inference time. Learn more in our #ICML26-accepted paper: https://neclab.eu/research-groups/system-platform-for-iot-and-ai/intelligent-software-system/publications#publications-2203 #NECLabs

ALT Our latest work introduces compositional steering tokens for controlling multiple #LLM behaviors. Encoding behaviors as input tokens enables robust zero-shot composition without modifying model internals. Learn more in our #ACL26-accepted paper: https://neclab.eu/research-groups/human-centric-ai/publications#publications-2201 #NECLabs

ALT We recently hosted an insightful talk by @peterbhase, Postdoctoral Researcher at @Stanford, on faithful reasoning and controllable knowledge in #LLMs. To learn more, watch https://youtu.be/P3S7cMBudBc. #NECLabs #AI

ALT Watch #NECLabs Senior Research Scientist @cc_hung_ explain how our modular explainable #differentialdiagnosis agent supports #clinicians by providing early diagnostic suggestions and enhancing accuracy and decision‑making. https://youtu.be/CIQthe4L4_w

ALT Watch Dr. Zarrar Yousaf share how #NECLabs Europe’s culture of #openinnovation, transparency and collaboration empowers researchers to explore new ideas and create technologies that benefit society. https://youtu.be/ONXacmsMnrY

ALT We hosted an insightful talk by @TomeEftimov on how benchmarking data can support more automated, transparent, and knowledge‑driven optimization and #ML. Watch here: https://youtu.be/oJFh4hbl4Kg. #NECLabs #AI

ALT We hosted an insightful talk by @federicobianchy on how #AI can go beyond search to actively discover new solutions to hard scientific problems. Watch here: https://youtu.be/k28Y2GVuFWE. #NECLabs #LLM #machinelearning

ALT We hosted an insightful talk by @gorjanradevski on compositional steering tokens, a new method for guiding #LLMs to follow multiple behaviors simultaneously by embedding behavioral instructions directly into input tokens. Watch here: https://youtu.be/_iZj_924Jwc. #NECLabs #AI