The Large Interferometer for Exoplanets (LIFE): a Space Mission for Mid-Infrared Nulling Interferometry
OPTICAL AND INFRARED INTERFEROMETRY AND IMAGING IX(2024)
Swiss Fed Inst Technol | Australian Natl Univ | Max Planck Inst Astron | Univ Liege | NASA | Katholieke Univ Leuven | Univ Bern | European Southern Observ | Univ Groningen | Free Univ Berlin | York Univ | CSIC | Aalto Univ | Georgia State Univ | Mines Paris PSL | Univ Warwick | Univ Roma Tor Vergata | CSIC INTA | Univ Toronto | Yale Univ | Cornell Univ | Univ Edinburgh | Stockholm Univ | Tech Univ Denmark | German Aerosp Ctr | OAW | Observ Paris | Univ Southern Queensland Montana State Univ | Univ Cambridge | McGill Univ | Aryabhatta Res Inst Observat Sci ARIES | MUSC | SRON Netherlands Inst Space Res | Univ Nacl Colombia | Network Life Detect | Univ Arizona | Ludwig Maximilians Univ Munchen | Ateleris GmbH | Univ Calif Santa Cruz | Leiden Observ | Univ Paris Saclay | Bayer Geoinst | Inst Radioastron Millimetr Tram | CCT La Plata CONICET UNLP | Univ Geneva | Blue Marble Space Inst Sci | Lab Astrophys Marseille | Univ Zurich | Christian Albrechts Univ Kiel | Univ Antofagasta | Univ Exeter | Inst Space & Astronaut Sci | Nagoya Univ | INAF Astron Observ Trieste | Univ Copenhagen | Univ Hertfordshire | Univ Calif Riverside | Univ Belgrade | Tech Univ Munich | Univ Duisburg Essen | Univ Cologne | INAF Osservatorio Astrofis Catania | Univ Tartu | TUDelft | Univ Chicago | Penn State Univ | Natl Inst Sci Educ & Res | Univ Grenoble Alpes | Campo Catino Observ | Univ Washington | CALTECH | Univ Michigan | ASTRON | Max Planck Inst Astrophys MPA | Purdue Univ | Inst Astrofis Canarias | Heliospace Corp | Slovak Acad Sci | ZAH Univ Heidelberg | Univ Cent Florida | Georg August Univ | STFC RAL | Polish Acad Sci | Vanderbilt Univ | Friedrich Schiller Univ | Delhi Technol Univ | UR Rao Satellite Ctr | Univ Lethbridge | Univ Tokyo | Harvard Univ
- Pretraining has recently greatly promoted the development of natural language processing (NLP)
- We show that M6 outperforms the baselines in multimodal downstream tasks, and the large M6 with 10 parameters can reach a better performance
- We propose a method called M6 that is able to process information of multiple modalities and perform both single-modal and cross-modal understanding and generation
- The model is scaled to large model with 10 billion parameters with sophisticated deployment, and the 10 -parameter M6-large is the largest pretrained model in Chinese
- Experimental results show that our proposed M6 outperforms the baseline in a number of downstream tasks concerning both single modality and multiple modalities We will continue the pretraining of extremely large models by increasing data to explore the limit of its performance
