China data and evidence directory / model

AI eval claim assurance

Name: AI eval claim assurance
Creator: Palimpsest
License: https://github.com/beepboop2025/palimpsest/blob/main/LICENSE

Dataset eval-assurance · Assessed 2026-10-08T12:47:59Z

Deterministic claim-by-claim audit of integrity, exact-prompt precommitment, raw-response recomputation, pipeline reproducibility, statistical design, independent human construct validation and unaffiliated replication. Publishes a claim ceiling instead of one composite quality score.

Within its collection window at this assessment; check the reading's reference period before use.

Current JSON

Open research view · Read the method

Evidence state: fresh
Reading timestamp: 2026-10-08T08:23:00.089388Z
Collection target: P1D
Sources: Palimpsest public eval registry, protocols, readings and transcripts
Geography: GLOBAL
History: No public history distribution
summary.checks: 11
summary.pass: 7
summary.partial: 1
summary.pending: 1
summary.open: 1
summary.fail: 1

Reuse terms: MIT. Source rights remain attached to observations.

How to use this evidence

The reading timestamp records the instrument's evidence clock. The economic reference period or event date is defined inside the source record. A recent collection does not turn an old observation into a current one. Counts describe this dataset's coverage, not all activity in China.

Cite this dataset

Palimpsest. AI eval claim assurance. Reading: 2026-10-08T08:23:00.089388Z. https://www.palimpsest.info/datasets/eval-assurance/

When citing a number, include the exact JSON or CSV record, its reference period, units and upstream source. Retain a copy of the reading; a latest URL changes as new editions are published.

Detailed economic tables · China evidence observatory · API and MCP access · Report a correction