FAR single-author machine-audited paper v2

Immutable public paper release for source commit 30dc37f. This version adds the post-hoc, zero-model selective-revision feasibility audit and preserves the frozen claim boundaries. It is machine-audited development evidence, not human review, human IAA, publication gold, external blindness, semantic correctness, a deployable selector, prospective calibration, causal policy utility, or strict submission readiness. The portable archive includes a standalone verifier and SOURCE.lock; verify with system Python in isolated mode.

FAR no-human TMLR paper release v1

Portable, independently verifiable release of the active FAR TMLR paper profile at commit 434414f. Includes the 15-page paper PDF, source lock, readiness v4 evidence, wheel, sdist, SBOM, benchmark validation, and secret scan inside a deterministic archive. The sidecar verifier uses only the Python standard library and is intended to run with python3 -I. Boundaries: no human review, adjudication, or IAA; no publication-grade human gold; no external blind test; no strict-submission readiness. P12 is a post-hoc lexical revision-trace fidelity audit with zero model calls and no held-out/test access.

FAR diagnostic artifacts v2

Immutable successor to artifacts-v1. Adds the P11 revision-delta evaluation audit and refreshed frozen diagnostic reports, including the post-hoc Mistral/Gemma/Llama typed-vs-untyped sensitivity analysis. No model calls were made for this refresh; human review, human adjudication, human IAA, publication gold, and external blindness remain unavailable. artifacts-v1 is unchanged.

FAR diagnostic artifacts v1

Frozen P0-P6 diagnostic payload. Verify the archive and its 336-file inventory against far/data/diagnostics-v1.json before use. This release contains machine-audited development diagnostics and P6 machine prelabels; it is not human gold, held-out evidence, or publication-grade annotation.