NEC Laboratories Europe compared encoders trained with and without Matryoshka Representation Learning (MRL) across 24 retrieval and classification tasks.
Public source
Publisher name
Public post
What if one of the most widely used techniques for compressing text embeddings isn't necessary for most applications? Text embeddings power AI applications ranging from…
Company
NEC Laboratories Europe
Orchestrating a brighter world
- Industry
- Research Services
- Location
- Heidelberg, DE
- Company size
- 51–200 employees
About NEC Laboratories Europe
NEC Laboratories Europe is based in Heidelberg, a city of science and culture, located in southwest Germany. Our research and development contributes to the NEC Group’s purpose of creating a more sustainable world in which everyone has the chance to reach their full potential. We work closely with the European Union and research partners from around the world to help transform science and technology in the areas of AI, security and privacy, digital health and information and communications technology. NEC Laboratories Europe has an internationally diverse workforce of over 100 employees from more than 25 different countries. Our researchers choose their projects, lead scientific discovery and frequently publish their findings at top, peer-reviewed scientific conferences. NEC Laboratories Europe R&D is active in five major technology areas: Generative and agentic AI, AI and IoT platforms, Life sciences, Security, and 6G networks.
See moreLatest activity
Latest activity from NEC Laboratories Europe
3 signals
Research & Knowledge
NEC Laboratories Europe collaborated with the University of Mannheim on the research paper titled To MRL or not to MRL.
Research & Knowledge
NEC Laboratories Europe published a paper titled OBLIVION: Self-Adaptive Agentic Memory Control through Decay-Driven Activation by Ashish Rana, Chia-Chien H., Qumeng Sun, Julian Kunkel and Carolin Lawrence, accepted at EMNLP 2026, in collaboration with The University of Göttingen.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
NEC Laboratories America, Inc.
NEC Laboratories America, Inc. published a paper titled Representation Interventions Enable Lifelong Knowledge Memory Control in LLMs on July 8, 2026.
Research & Knowledge
Google DeepMind
Google DeepMind tested 27 candidate languages across 12 diverse models, two of them analyzed through pretraining at EMNLP 2026.
Research & Knowledge
Label Your Data
Label Your Data collected and annotated 15,000 images for human pose estimation using skeleton annotation for Nex, contributing to a 12% improvement in model accuracy.
Research & Knowledge
Target
Target conducted research on improving RAG results by including bi- and cross-encoders in the process and using completely synthetic training data.
Research & Knowledge
Aleph Alpha