NEC Laboratories Europe GmbH’s Post

NEC Laboratories Europe GmbH is proud to be sponsoring the European Big Data Value Forum (EBDVF) 2024, which kicks off next Wednesday in Budapest, Hungary. The three-day conference will feature insightful sessions and workshops on the way forward for big data and data spaces, highlight how businesses can harness the power of trustworthy AI and discuss the role of high-performance computing as an enabler for digital transformation. As a highlight of the event, our Program Manager Dr. Roberto González Sánchez will be giving a talk on the potential real-world applications of generative AI agents. You can learn about our AI work at https://neclab.eu. #NECLabs #LLM #generativeAI #machinelearning #artificialintelligence #EBDVF24

NEC Laboratories Europe GmbH were privileged to participate in the 11th Heidelberg Laureate Forum yesterday in Heidelberg, Germany. Our Senior Researcher Dr. Raman Siarheyeu presented our current research on accelerating therapeutic cancer vaccines and universal vaccines for infectious diseases with AI. Organized by the Heidelberg Laureate Forum Foundation, the HLF is a networking conference that brings together exceptional young researchers with the recipients of the most renowned prizes in the fields of mathematics and computer science. For more information on #NECLabs biomedical AI research, visit: https://lnkd.in/ee9DUVw8. #HLF24 #AI #biomedical #computerscience #mathematics

NEC cotomi generative AI takes another impressive step forward by integrating LLM Explainer, the innovative NLP technology developed by NEC Laboratories Europe GmbH to detect hallucinations and anomalies created by LLMs. By comparing LLM output to input documentation, LLM Explainer allows the user to verify the correctness of LLM-generated output by linking to relevant input documentation. This helps overcome the need for time-consuming manual checking of LLM-generated text and paves the way for wider adoption of LLM technology in sensitive use cases. LLM Explainer can be used with commonly available LLMs. To learn more, read the press release https://lnkd.in/eYGFfUYU. #NECLabs #NECcotomi #LLM #GenerativeAI #NLP

While state-of-the-art language models (LMs) deliver near-human performance levels on a wide range of problems, they sometimes struggle with noisy, real-world data. This drop in performance, caused by even minor and irrelevant text changes, poses a significant problem for downstream applications. To address this, we investigated prototype-based networks (PBNs), which classify based on similarity to prototypical examples. Our research shows that a PBN architectural variation of vanilla LMs enhances robustness more effectively than traditional LMs and offers inherent interpretability, providing both transparency and reliability. Additionally, PBN robustness is influenced by hyperparameters such as the number of prototypes and tight clustering conditions during training. The work is a collaboration with the USC Information Sciences Institute, USC Thomas Lord Department of Computer Science, Vrije Universiteit Amsterdam (VU Amsterdam) and Ss. Cyril and Methodius University in Skopje / Универзитет „Св. Кирил и Методиј“ во Скопје. To learn more, read our paper, “Robust Text Classification: Analyzing Prototype-Based Networks” by Zhivar Sourati, Darshan Deshpande, Filip Ilievski, Kiril Gashteovski and Sascha Saralajew, accepted at Findings of EMNLP 2024. https://lnkd.in/eim4tSzy. #NECLabs #NLP #languagemodels #EMNLP24

It’s a wrap! Thanks to Avaloq for an amazing 2024 community conference, where Avaloq discussed its roadmap for the future of wealth management, including the role of AI in banking. We are privileged to be part of this journey.  It also was an ideal setting to showcase our Safe LLM Framework demo to Avaloq customers. The framework uses human-centric AI to identify and highlight inaccuracies and hallucinations often present in LLM-generated information. The application can be used with any LLM to accelerate content creation where accuracy is essential. To learn more about NEC’s work on developing safer large language models, visit https://lnkd.in/erpm_tKG. #NECLabs #GenerativeAI #GenAIagents #Banking #AvaloqCommunity

We are excited to be joining Avaloq and their customers – some of Europe’s leading banks and wealth managers – for the annual Avaloq Community Conference, taking place tomorrow in Zurich, Switzerland. During the event, our General Manager Roberto Baldessari and Avaloq’s Chief Banking Operations Officer Roberto Viganò will introduce the first solution in the banking industry based on generative AI for corporate actions processing. #NECLabs #GenerativeAI #Banking #AvaloqCommunity

We're #hiring! We are looking for a Security Engineer to join our Research Lab in Heidelberg! Apply today or share this post with your network. #cryptography #softwaredevelopment #systemssecurity https://lnkd.in/eKhfCJ47

Large language models (LLMs) are revolutionizing society and industry and transforming how we interact with intelligent systems. They allow us to enhance productivity across industries by generating fluent text, learning new tasks quickly and assisting with programming. However, LLMs are prone to hallucinations and may provide false information in their answers. They struggle with complex reasoning and reflect biases in their training data. To harness the benefits of LLMs while minimizing risks, we have developed the Safe LLM Framework, which includes tools like LLM Explainer and Fact Checker. LLM Explainer links generated text back to original sources, enabling users to verify the accuracy of the information. Fact Checker compares outputs to trustworthy sources and provides reasoning paths, ensuring the reliability of the generated content. By implementing these strategies and keeping humans in control, we can leverage the immense potential of LLMs for a brighter, more equitable future. To learn more, read our case study, “Towards Safer Large Language Models”: https://lnkd.in/erpm_tKG. #NECLabs #artificialintelligence #LLMs #generativeAI #explainableAI #responsibleAI