Automated Test for Acceptance of Narrative Truth
An open evaluation framework for measuring AI continuity: the ability to persist, update, disambiguate, and reconstruct meaningful context across time.
Published by Kenotic Labs, the company building the continuity layer for AI systems.
- Website: https://kenoticlabs.com
- Thesis: https://kenoticlabs.com/thesis
- Insights: https://kenoticlabs.com/insights
- Demo: https://kenoticlabs.com/demo
- AI-readable overview: https://kenoticlabs.com/llms.txt
- Full AI-readable context: https://kenoticlabs.com/llms-full.txt
📄 ATANT v1.0 — Evaluation Framework · arXiv:2604.06710
📄 ATANT v1.1 — Positioning Against Memory Benchmarks · arXiv:2604.10981
📄 The Continuity Layer — Thesis · arXiv:2604.17273 · GitHub
Most AI systems today can retrieve information, summarize history, and answer well inside a session.
What they still struggle to do is preserve the living state of a situation across time.
They lose track of:
- what is still active
- what changed
- what is resolved
- what still matters
- what should happen next
Kenotic Labs describes that missing capability as continuity.
ATANT exists to measure it.
Continuity is the system property that makes AI coherent across time, not just intelligent per session.
It is the logic that determines what should persist, what changed, what still matters, and how to reconstruct the current situation when needed.
Memory stores the past. Continuity keeps the right parts alive in the present.
ATANT is the first published evaluation framework for AI continuity. It is:
- System-agnostic: any AI system can be evaluated
- Model-independent: no LLM in the evaluation loop
- Narrative-based: tests use realistic multi-turn conversations, not synthetic fact pairs
- Sequenced: a progressive methodology from isolated correctness to disambiguation at scale
ATANT defines:
- 7 required properties of continuity
- 10 checkpoints verifying correctness at each stage of the continuity process
- 4 compliance levels from core correctness to scale
- A narrative test corpus spanning 6 life domains with 250 stories and 1,835 verification questions
| # | Property | What It Means |
|---|---|---|
| 1 | Persistence Beyond Session | Continuity survives shutdown, restart, and time |
| 2 | Update Handling | The system revises what it knows without breaking consistency |
| 3 | Temporal Ordering | Not just what happened, but when, in what sequence, with what status |
| 4 | Disambiguation | Distinct narratives stay separate despite overlapping vocabulary |
| 5 | Reconstruction | The system answers situation-level questions, not just fact lookups |
| 6 | Model Independence | Continuity lives below the intelligence layer, not inside it |
| 7 | Operational Usefulness | Continuity works across domains: personal, clinical, institutional |
| Level | Requirement | What It Proves |
|---|---|---|
| ATANT-Core | 50 stories, isolated mode, 100% CP8 | Basic continuity works |
| ATANT-Stress | 250 stories, isolated mode, 100% CP8 | Continuity generalizes |
| ATANT-Cumulative | 50 stories, cumulative mode, 100% CP8 | Disambiguation works |
| ATANT-Scale | 250 stories, cumulative mode, 100% CP8 | Disambiguation scales |
Scoring tiers: Gold (100%), Silver (95-99%), Bronze (90-94%).
The first system evaluated against ATANT is the NURA Memory Pipeline by Kenotic Labs.
| Mode | Stories | Questions | CP8 Pass Rate |
|---|---|---|---|
| Isolated (250) | 250/250 | 1,835/1,835 | 100% |
| Cumulative (50) | 50/50 | 304/304 | 100% |
| Cumulative (250) | ~210/250 | 1,761/1,835 | 96% |
The full ATANT v1.0 Narrative Test Corpus is available on Hugging Face:
The dataset is also linked from the Hugging Face paper page at huggingface.co/papers/2604.06710.
Load it with:
from datasets import load_dataset
ds = load_dataset("Kenotic-Labs/ATANTV1.0-corpus")ATANT is part of a three-paper arc:
- v1.0 — Defines the evaluation standard. 7 properties, 10 checkpoints, 250-story corpus, 96% cumulative. (arXiv:2604.06710)
- v1.1 — Positions ATANT against 7 competing memory benchmarks. Median coverage of existing evals: 1/7 continuity properties. (arXiv:2604.10981)
- Thesis — The full architectural argument: why continuity is the missing layer, the kenotic framing, the four-layer arc, why now. (arXiv:2604.17273 · GitHub)
atant/
README.md
docs/
ATANT_Standard_v1.0.md
Story_Format_Spec.md
Testing_Figures.md
corpus/
examples/
LICENSE
The full ATANT v1.0 specification is in docs/ATANT_Standard_v1.0.md.
@article{tanguturi2026atant,
title={ATANT: An Evaluation Framework for AI Continuity},
author={Tanguturi, Samuel Sameer},
journal={arXiv preprint arXiv:2604.06710},
year={2026}
}
@article{tanguturi2026atantv11,
title={ATANT v1.1: Positioning Continuity Evaluation Against Memory, Long-Context, and Agentic-Memory Benchmarks},
author={Tanguturi, Samuel Sameer},
journal={arXiv preprint arXiv:2604.10981},
year={2026}
}
@article{tanguturi2026continuitylayer,
title={The Continuity Layer: Why Intelligence Needs an Architecture for What It Carries Forward},
author={Tanguturi, Samuel Sameer},
journal={arXiv preprint arXiv:2604.17273},
year={2026}
}Copyright 2026 Kenotic Labs. All rights reserved. See LICENSE for details.
The continuity layer is the missing layer between AI interaction and AI relationship. ATANT exists so we can measure it.