Field note /
More context is not the same as cleaner evidence
Controlled review work showed that semantic judgments can become sharper without becoming more discriminating, and that agent-authored framing can move the result.
How the claim changed
Belief
If semantic review receives richer implementation context, uncertainty should fall and defect discrimination should improve together.
Evidence
Richer context produced substantially more decisive Jev outputs in a controlled review experiment.
Contradiction
The richer condition did not reliably separate the defective commit from the corrected commit. Follow-up scenarios also showed that unlabeled agent opinion, prior semantic answers, and proposal framing could shift judgments, while verbatim deterministic tool output moved them on a different footing.
Correction
Treat semantic context as a contract: include primary artifacts and scoped deterministic evidence, label claims as claims, exclude the choosing agent’s verdict and prior Jev answers by default, and freeze serialization/order when results are compared.
Current claim boundary
These findings establish context sensitivity and contamination risk for the tested case and model; they do not establish a universal effect size or a general semantic-provider ranking.