Frozen random-weight attention transformers can emulate kernel regression and approximate Hölder functions at minimax-optimal rates, with soft prompts constructed by solving linear systems.
neurips.cc/paper_files/paper/2008/file/0efe32849d230d7f53049ddc4a4b0c60-Paper.pdf
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Training-Free Universal Approximation by Prompting Random Transformers
Frozen random-weight attention transformers can emulate kernel regression and approximate Hölder functions at minimax-optimal rates, with soft prompts constructed by solving linear systems.