DeepSeek's Liang Wenfeng Co-Authored Paper Claims Best Paper Award at ACL 2025

2 week ago / Read about 0 minute

Author：小编

At the prestigious ACL 2025 Awards Ceremony, a paper co-authored by Liang Wenfeng of DeepSeek, alongside collaborators from Peking University, emerged victorious, claiming the Best Paper Award. The groundbreaking work introduces the Native Sparse Attention (NSA) mechanism, which revolutionizes the processing of long texts by boosting speeds an impressive 11-fold through a synergistic combination of algorithmic and hardware optimizations. This innovative approach surpasses traditional full attention models, setting a new benchmark in the field.

Previous page：Shenzhen Coocaa Receives Pioneering AI Hardware In...

Next page：Zuckerberg: AI Propels Growth in Meta's Advertisin...

Return to List

Hot Reading

2 day ago

Trump's Anti-Science Agenda Is Massively Hampering His Plans for AI, Experts Warn

2 day ago

Is GPT-5 really worse than GPT-4o? Ars puts them to the test.

2 day ago

New Research Finds That ChatGPT Secretly Has a Deep Anti-Human Bias

2 day ago

Anthropic says some Claude models can now end ‘harmful or abusive’ conversations