The claim
A86 §9: “There is already an active literature on semantic and goal-oriented communication, including semantic rate–distortion and receiver side information. We should build on that rather than present task-sensitive information preservation as an entirely new invention.” It named a survey, Chai et al. (rate–distortion–perception with encoder and decoder side information) and Zhao et al. (2025), Semantic Rate-Distortion Theory with Applications, which “explicitly studies ambiguity, polysemy, and distortion of conditional semantic probability distributions. That comes close to our interest in preserving distributions over interpretations rather than one purported intrinsic meaning.” The narrowed contribution: “Apply information-theoretic preservation criteria to sheaf-structured pragmatic observations, and study how those structures change under sequential, context-dependent reconstruction.”
Where it went
C6 has none of these works and no concession. The manuscript’s §1 disclaims novelty for each ingredient (CLAIM-023), but its §5 presents task-relative rate–distortion and side information with Shannon, Blackwell, Torgersen and Wyner–Ziv only. Granted, because A86’s own warning applies to the manuscript as it stands. The cited works were found by web search (A86’s cite tags); neither the transcript nor the record establishes their details.
The prior art, read
The three works the exchange conceded are now read:
Zhao et al. formalise preserving the posterior p(S|x), the distribution over interpretations that the manuscript wanted. Under KL divergence their semantic distortion is the information bottleneck; under total variation it bounds the lost value of every bounded-loss decision (THEORY-161; both are the reader’s derivations). The concession stands, and more strongly than the exchange knew. What the manuscript can still claim is the context-indexed, sheaf-structured observation, not task-relative preservation.
Sheaf-structured meaning predates the manuscript
By 2014, sheaf-structured meaning for discourse already existed: Abramsky and Sadrzadeh (LIT-841). Only the pragmatic and transport use of it remains for the manuscript to claim, which supports this claim’s narrowing.