What the paper is about: The authors show that an information-theory principle called information causality can single out exactly the set of quantum correlations in the simplest Bell scenario — the case where two distant parties each choose one of two measurements and each measurement has two outcomes. In plain terms, they prove that a limit on how much information a receiver can gain about a sender’s bits, even when the parties share exotic non-signaling resources, leads to the same boundary that quantum theory permits in this basic nonlocality test.

What the researchers did: They used a strengthened, “generalized” version of information causality that allows the sender’s two input bits to be correlated. They also designed a new communication protocol that sends the sender’s full classical record through a deliberately noisy binary channel and keeps the receiver’s entire transcript (the channel output plus any outcome from the shared resource). By choosing the input correlations and the channel’s transition probabilities carefully, and by examining the conditional mutual information between sender and receiver, they derived the Tsirelson–Landau–Masanes (TLM) criterion — a known mathematical condition that exactly describes all quantum-realizable correlators in this simplest Bell scenario.

How it works at a high level: Information causality limits how much information can be transmitted through a classical channel, even when parties have access to nonlocal correlations. By letting the sender’s inputs be correlated and by not compressing the receiver’s information into a single guess, the authors avoided information loss that earlier tests suffered. They also tuned the noisy channel (including taking a vanishing capacity limit) so that the small transmitted information reveals precise constraints on the observable two-party correlators. Applying this approach yields the TLM inequality directly from the information principle.