Paper 9 public benchmark validation transcript
Date: 2026-05-03

Scope
-----
Validate the public benchmark evidence for the receipted-actions/adtech paper.
This transcript covers public deterministic synthetic data only. It does not
include design-partner or other partner traffic.

Determinism hardening
---------------------
Issue found during validation: event and verdict files replayed byte-for-byte,
but the generated manifest drifted because it carried a wall-clock timestamp.
The internal benchmark generator was updated to pin the canonical generated-at
value for v1 so that README, events, verdicts, and manifest replay byte-for-byte.

Targeted internal validation
----------------------------
Result:
  60 passed, 1 warning

Canonical release summary
-------------------------
  scale: 10k
  events_observed: 10150
  verdicts_emitted: 10150
  verdict_mix: allow=8694, hold=135, block=1321
  blocked_dollar_by_currency: USD=7905.76

Byte-for-byte replay
--------------------
Result:
  byte_for_byte_replay_ok

Research import validation
--------------------------
Imported path:
  3-proof/3d-evidence/adtech/09-receipted-actions/benchmark_v1_10k/

Command:
  sha256sum -c SHA256SUMS

Result:
  README.md: OK
  benchmark_v1_10k.events.jsonl: OK
  benchmark_v1_10k.verdicts.jsonl: OK
  benchmark_v1_10k.manifest.json: OK

Line counts:
  10150 benchmark_v1_10k.events.jsonl
  10150 benchmark_v1_10k.verdicts.jsonl
  20300 total

Interpretation
--------------
The public benchmark is a real replayable experiment over deterministic
synthetic rewarded-action claims. It supports a publishable systems claim about
receipted-action envelopes, verdict replay, reason-code coverage, and appeal
shape. It does not support a live customer fraud-lift claim; that remains gated
on partner-approved aggregate evidence.

Public-release hardening
------------------------
The public release intentionally omits internal repo names, operational source
paths, live review endpoints, and exact private tuning mixes. The released
artifact remains checksum-verifiable and quantitatively consistent with the
public systems claim.
