The paper proposes injecting error-correcting-code-style tokens into LLM prompts to improve associative recall, but offers no measurements to support the claim.
Candès and Terence Tao
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Hypertokens: Holographic Associative Memory in Tokenized LLMs
The paper proposes injecting error-correcting-code-style tokens into LLM prompts to improve associative recall, but offers no measurements to support the claim.