A weak Hellinger inequality for noisy Boolean channels

Polona Durcik, Marco Fraccaroli, Joris Roos

The paper proves the weak Hellinger inequality for a noisy Boolean channel with a one-bit summary at each end. A central step is a three-parameter inequality proved with Bernstein-polynomial bounds and finite positivity checks.

What was already known

Chen, Gohari and Nair posed a broader Hellinger conjecture and developed the finite-dimensional reduction used in this work.

AI involvement AI-centered

The authors credit OpenAI GPT models with substantial work on the proof of the central three-point inequality during extended exchanges. They wrote the manuscript and Lean theorem statements. Codex generated the Lean proofs under their supervision.

Importance Notable

The theorem establishes a two-summary special case of the Hellinger problem through a concrete inequality that can be checked in Lean.

Technical verification Formal (Lean)

The paper links the Lean statement for the weak inequality and distinguishes it from the separate statement for its three-point inequality.

Formal checking

Checked the weak Hellinger theorem in the pinned Lean source.

Checker
Lean 4 kernel
Scope
The paper proves the weak Hellinger inequality for a noisy Boolean channel with a one-bit summary at each end. A central step is a three-parameter inequality proved with Bernstein-polynomial bounds and finite positivity checks.
Basis
Reported by the linked checker

Human understanding Few

The authors explain their reduction and the Bernstein-polynomial method. The disclosure leaves independent human reconstruction of the central AI-assisted positivity proof undocumented.

Other useful sources

Correction and publication history