Eagle and Finch enhance RWKV with matrix-valued states and dynamic recurrence, trained on a 1.12-trillion-token multilingual corpus, and report competitive performance on standard benchmarks.
Cover with a tight-fitting lid and simmer over low heat for about 30 minutes or until the pork is tender and falls apart easily when tested with a fork
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence
Eagle and Finch enhance RWKV with matrix-valued states and dynamic recurrence, trained on a 1.12-trillion-token multilingual corpus, and report competitive performance on standard benchmarks.