Primal-Attention: Self-attention through Asymmetric Kernel SVD in Primal Representation

Open paper
Benchmark Model Rank Results
d4rl-on-d4rlPrimal.+DT#2Average Reward: 77.5
language-modelling-on-wikitext-103Primal.+Trans.#66Test perplexity: 31.0

Rank counts only results with a code link.