This paper studies a laboratory version of a thought experiment about black holes and information. The authors adapt the Hayden–Preskill recovery idea so it works at finite temperature and so the injected piece of information is related to the system that scrambles it. They test the modified protocol with a model known to scramble quantum information quickly, and they run a proof‑of‑principle experiment on an IBM superconducting quantum processor.

In the original Hayden–Preskill setup, a small quantum state (the “diary”) is thrown into an already entangled system and, after scrambling, an outside observer can recover the diary from emitted radiation. The authors change the injection step: instead of appending the diary to the system, they swap it with one qubit inside the system using a SWAP gate. This SWAP method means the same Hamiltonian can be used to prepare the initial state and to drive the later scrambling, avoiding ambiguities that arise if the diary is simply appended. They also work at finite temperature by preparing a thermofield double (TFD) state — a way to represent thermal entanglement between two systems.

To quantify recovery they use two standard measures: the postselection probability P (how often the decoding step succeeds) and the conditional fidelity F (how well the diary is recovered when decoding succeeds). Using the Sachdev–Ye–Kitaev (SYK) model — a many‑body Hamiltonian known for fast scrambling — they derive exact early‑time values and late‑time estimates for P and F. The late‑time formulas assume uniform operator spreading, and the authors report that these analytic estimates agree well with their numerical simulations. A key theoretical finding is that both P and F scale roughly in proportion to temperature, which the authors trace to reduced entanglement in the low‑temperature initial state. They also show the SWAP injection produces an extra temperature factor (called ξβ in the paper) not present in conventional appended decoders.