An interactive QA protocol lets a small LLM recover 23-72% of the capability gap to a large model using 10 yes/no answers, achieving compression ratios of 0.0006-0.004, over 100x better than prior LLM compression methods.
If the score ≥ 7, the protocol accepts the current answer and early-stops
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Haiku to Opus in Just 10 bits: LLMs Unlock Massive Compression Gains
An interactive QA protocol lets a small LLM recover 23-72% of the capability gap to a large model using 10 yes/no answers, achieving compression ratios of 0.0006-0.004, over 100x better than prior LLM compression methods.