Search
DeepSeek's Revolutionary Mathematical Model Breaks Records: Compact 7B Model Independently Uncovers Skills Eluding 671B Models
14 hour ago /
Read about 0 minute
Author：小编
DeepSeek has unveiled its latest model, DeepSeek-Prover-V2, which is dedicated to mathematical theorem proving and has achieved groundbreaking results across several high-difficulty benchmarks. In the prestigious Putnam test, DeepSeek-Prover-V2 solved an impressive 49 problems, vastly outperforming the current leader, Kimina-Prover, which managed only 10 problems. Kimina-Prover is a collaborative effort between Kimi and the esteemed AIME2024 champion team Numina. The exceptional performance of DeepSeek-Prover-V2 underscores its formidable prowess in the realm of mathematical theorem proving, even with a relatively compact 7B parameter count, showcasing skills that elude much larger 671B models.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
