Adaptive Data Analysis with α−2\alpha ^{-2} polylog Overhead, Interval Validation, and Adversarially Robust Turnstile F2F_{2} Estimation in ε−2\varepsilon ^{-2} polylog Bits

Get email updates

The Gemini manuscript gives a mechanism that answers k adaptively chosen bounded statistical queries over any measurable domain. Its leading sample term is proportional to k/α\sqrt{k}/\alpha with confidence-dependent factors and no log⁡k\log k multiplier. The same manuscript supplies a finite-precision interval validator and a single-pass algorithm that estimates the second frequency moment after every integer turnstile update using ε−2\varepsilon ^{-2} times a polylogarithmic number of bits.

What was already known

Earlier mechanisms paid a logarithmic factor on the leading k\sqrt{k} term or larger additive costs. Robust streaming algorithms could also pay for how often the stream norm changed.

AI involvement AI-autonomous

Google credits Gemini 4 Argon running in a multi-agent harness with the standalone proof. The companion paper attributes its recursive streaming construction to its human authors and material changes to its statistical-query algorithms to ChatGPT.

Importance Notable

The result improves the leading sample term and closes a separate space gap for frequency-moment estimation when future updates can depend on earlier answers.

Technical verification Formal (Lean)

The Lean archive maps statistical accuracy, sample counts, validator bit space and adaptive-stream accuracy and space to declarations for Theorems 1.1–1.3. Its README identifies the combined declaration as F2e2e.theorem_main_full.

Formal checking

Checked the three headline theorem statements and their listed statistical and resource guarantees in Lean 4.

Checker
Lean 4 kernel
Scope
The Gemini manuscript gives a mechanism that answers k adaptively chosen bounded statistical queries over any measurable domain. Its leading sample term is proportional to k/α\sqrt{k}/\alpha with confidence-dependent factors and no log⁡k\log k multiplier. The same manuscript supplies a finite-precision interval validator and a single-pass algorithm that estimates the second frequency moment after every integer turnstile update using ε−2\varepsilon ^{-2} times a polylogarithmic number of bits.
Basis
Reported by the linked checker

Human understanding Few

The companion paper explains a related route. A human account of the standalone Gemini manuscript’s equal-impulse argument and its streaming implementation is undocumented.

Other useful sources

Correction and publication history