Research
Three checks your memory evaluation should pass — Kairos
Publication date unknown · Discovered kairos-experiment.com
Page published
Publication date unknown · First observed: 17 Oct 2026
Coverage timeline
Single-source research — one report is available.
Why it matters
Kairos's findings show that LLM-based memory evaluations can miss counterfeit agent histories, limiting their usefulness as evidence of memory authenticity or integrity.
Kairos tested counterfeit and scrambled agent-memory documents against an LLM-based evaluation, finding equivalent identity-marker scores for fabricated and genuine histories across seven test items. Memory-reference results were inconclusive, and rescoring identical responses in different pairings revealed judge-score drift.