Joseph Schindler, Philipp Strasberg, Niklas Galke, Andreas Winter, Michael G. Jabbour (2025), arXiv preprint — ARXIV-2503.15612
Key takeaways
- A single coarse-grained entropy, S^τ_M(ρ) = S(τ) − D_M(ρ‖τ), which combines a measurement M (observational entropy) with a max-entropy prior τ set by physical constraints (Jaynes), recovers most entropies used in physics as special cases and admits general second-law theorems.
Seeded from the abstract and a skim, before the reading in NOTE-043. What follows is what the work says about itself.
The authors define a coarse-grained entropy that merges two traditions: measurement-based observational entropy and Jaynes-style maximum-entropy coarse-graining. The bridge is to treat physical constraints as information-theoretic priors. The new quantity contains most physically relevant entropies as special cases, supports new entropy-increase theorems with links to thermodynamics, and removes pathologies that traditional observational entropy has in infinite dimensions. They illustrate its dynamics in a quantum random-matrix model and a classical hard-sphere gas, and propose it as a basis for a general approach to statistical mechanics.
Standing in the record
Filed from the survey of 2026-09-25 of work the anthology set aside as out of scope (tier C): 365 seconds of active reading over 3 sessions in the papers-feed tracker. It was filed Deferred, unread. NOTE-043 is the close reading of 2026-09-25, and it placed the work: Active — worth reading as the reference definition of prior-generalised observational entropy and for its catalogue of special cases (§V, Table II). The unification is definitional and its bounds are correct. The second-law programme is a framework plus partial sufficient conditions, not a proved second law for any concrete system.
When seeded, before the reading in NOTE-043, the entry gave a priority for a deeper reading: low — A long technical paper whose central definition and scope the skim already captures. The owner’s time is moderate (t=365 s). It has no bearing on anthology practice.
What the seeding asked a deeper reading to check (NOTE-043 is that reading):
- This is foundations of statistical mechanics. Its tie to the anthology is the relative-entropy/prior structure, which resembles coarse-grained information measures used in learning theory. There is no ML content.
- A deeper reading should check the precise statements of the second laws in §VII (which regimes, which assumptions on τ) and whether the paper has since been published.
Access when seeded: I read the arXiv abstract page and the full v1 PDF (35 pp.). That covered the introduction, table of contents, §IV (Definition), the list of special cases in §V, and §X (Conclusions). I only skimmed the headings of §VI–IX and the appendices. arXiv lists no journal reference. A Crossref search found no published version, and OpenAlex was rate-limited, so publication status is unverified.