MathLabs

Worked solution: Huang's signed hypercube adjacency matrix proof of the sensitivity conjecture (2019)

Step 7 of 7: Closing the loop: the sensitivity conjecture is settled
In plain words

One more ingredient, proved earlier by other authors, finishes the job: block sensitivity bs(f)bs(f) is never much bigger than deg⁡(f)\deg(f). Chaining that fact together with Huang's new bound s(f)≥deg⁡(f)s(f) \ge \sqrt{\deg(f)} produces a direct polynomial link between bs(f)bs(f) and s(f)s(f), exactly what Nisan and Szegedy had asked for in 1992.

bs(f)≤s(f)4bs(f) \le s(f)^4
Detailed analysis

Nisan and Szegedy (1992) had already shown bs(f)≤deg⁡(f)2bs(f) \le \deg(f)^2 up to a constant factor, later sharpened by Tal (2013) to exactly bs(f)≤deg⁡(f)2bs(f) \le \deg(f)^2. Squaring Theorem 1.4 from Step 6 gives deg⁡(f)≤s(f)2\deg(f) \le s(f)^2, and substituting into Tal's bound gives bs(f)≤s(f)4bs(f) \le s(f)^4 for every Boolean function ff. This is precisely the polynomial relationship the Sensitivity Conjecture asked for (with the explicit constant C=4C=4), so Huang's short spectral argument -- built from nothing more than one clever matrix, one classical eigenvalue-interlacing theorem, and one elementary row-sum estimate -- closes a problem that had resisted attack for close to thirty years.

Terms in this step
Block sensitivity bs(f)bs(f)
A relaxation of sensitivity: the largest number of pairwise disjoint blocks of coordinates that can each individually be flipped to change the value of ff at some input.