Research

validatedPublic libraryABIS program

A research librarybuilt like a live program,not a dead archive.

Papers, findings studies, validation notes, and applied scenarios for deterministic behavioral measurement.

Browse whitepapers Read research notes Open ABIS program

Library map

Papers, findings, validation notes, and use cases now read as one system.

The public research layer now shows the ABIS program as an active publication run: flagship claims, dated benchmark releases, lighter notes, and operational scenarios all feeding the same narrative.

11 research papers4 live notes1 dated findings

ABIS

core program

Validated

evidence mode

Public

library access

Start with the ABIS program

Latest findings study

Frontier Stability Snapshot - March 2026

A public release summarizing how the current ABIS paper program frames frontier model stability, drift, and release-to-release consistency.

Open 10 Mar 2026

Latest note

Paper Program Update

The library now tracks the current ABIS paper set directly on-site so the research archive reads like an active program rather than a placeholder page.

Read 10 Mar 2026

research papers

research notes

findings studies

applied scenarios

What lives here

Whitepapers that formalize the main ABIS claims.
Dated findings studies that package benchmark results into public releases.
Research notes that keep the program active between papers.
Validation and disclosure pages that explain how evidence is handled.

Current program

The strongest thread in the current library is ABIS: deterministic behavioral measurement for AI systems, with an emphasis on silent model change, behavioral stability, and operational governance.

How to read the library

Start with the ABIS program page for the overall framing.
Read the whitepapers for the main claims and evidence.
Use the drift-monitor section for dated findings and benchmark releases.
Follow research notes for lighter updates and release context.

Open ABIS

Browse Whitepapers

Read Research Notes

See Drift Monitor Studies

Program entry

ABIS

Open the flagship program page first if you want the measurement philosophy before the individual papers.

Open program

Disclosure layer

Validation

See how CIJ Labs frames evidence, publication boundaries, and what can be disclosed safely in public.

Read validation

Operational narrative

Applied scenarios

Follow the path from research claims to production governance, monitoring, and reliability use cases.

See scenarios

Program

ABIS

Overview of the flagship CIJ Labs program for deterministic behavioral measurement and live drift monitoring.

Open

Methodology

Validation

Evidence policy, disclosure boundaries, and the rules that make the public library trustworthy.

Open

Findings

Drift Monitor

Dated benchmark-style studies documenting what has been found in monitored model behavior.

Open

Application

Use Cases

Applied scenarios that show where the measurement layer fits into production AI operations.

Open

Research Papers

11 papers in the current program.

Longform papers anchor the research program and make the main claims legible without exposing sensitive implementation details.

View all papers

Published paper

Deterministic Detection of Large Language Model Behavioral Drift

A deterministic approach to detecting LLM behavioral drift using information-theoretic and mathematical text signatures, without training data, learned monitors, or baseline distributions.

10 Mar 2026PDF available

Research manuscript

Dual-Tier Feature Hierarchy in Behavioral Drift Detection

Feature-importance analysis across two training regimes suggests a hierarchy of universal indicators and domain-specific specialists in behavioral drift detection.

5 Mar 2026Summary live

Research manuscript

A Behavioral Geography of Large Language Models

A 29-model mapping study that treats LLM behavior as a measurable geography in high-dimensional feature space rather than as a simple ranking problem.

22 Feb 2026Summary live

Research manuscript

Catastrophic Fragility in Sequential Drift Detection

A case study showing catastrophic training collapse in a Transformer-based drift detector, contrasted with successful XGBoost performance on the same task.

18 Jan 2026Summary live

Research manuscript

Behavioral Rigidity Spectrum

A quantitative study of how resistant different frontier LLMs are to prompt-level correction once drift has been detected.

20 Oct 2025Summary live

Research manuscript

Closed-Loop Deterministic Correction of AI Behavioral Drift

A closed-loop system that detects behavioral drift and attempts escalating correction, revealing a boundary between correctable and architecturally irreversible change.

15 Aug 2025Summary live

Latest Notes

Fast-moving updates between papers.

Research notes keep the library active while manuscripts and public benchmark studies are being packaged.

Read notes

Research note

Paper Program Update

The library now tracks the current ABIS paper set directly on-site so the research archive reads like an active program rather than a placeholder page.

10 Mar 2026validated

Open note

Research note

Behavioral Geography Preview

The model-mapping paper introduces a useful framing shift: providers and versions can be thought of as occupying a measurable geography, not just a leaderboard rank.

6 Mar 2026validated

Open note

Research note

Correction Boundary Note

Closed-loop correction results are changing the program's posture from 'detect and fix' to 'detect, classify, and decide whether intervention is realistic.'

4 Mar 2026validated

Open note

Research note

Validation Cadence Update

Public note on weekly benchmark cadence and publication delay policy.

24 Feb 2026validated

Open note

Findings Studies

Dated releases built from monitored behavior.

These pages connect the paper program to concrete benchmark releases and public observations.

Browse findings

Frontier models

Frontier Stability Snapshot - March 2026

A public release summarizing how the current ABIS paper program frames frontier model stability, drift, and release-to-release consistency.

10 Mar 2026Stability and capability diverge more often than most model comparisons acknowledge.

Read findings

Applied Scenario

How the research maps into operations.

Use cases turn the research library into an operational narrative for teams who care about reliability and drift.

See scenarios

Applied AI

Production LLM Governance for Customer-Facing Systems

A practical scenario for using ABIS to monitor customer-facing model behavior when product quality depends on consistency, not just raw capability.

7 Mar 2026Earlier drift response

Read scenario

Core Themes

The wider CIJ Labs context is still here.

The library now makes papers and studies first-class, but the original thematic pages remain available for broader framing.

Theme

Behavioural modelling

The mathematical language used to fingerprint model behavior and measure structural change over time.

Open theme

Theme

Interpretability

Why deterministic receipts and measurement layers matter when models operate in opaque environments.

Open theme

Theme

System design

Organism-inspired architecture patterns for connecting measurement, alerting, and intervention loops.

Open theme

Theme

Decision systems

The wider CIJ Labs frame: how structured measurement supports real-world operational governance.

Open theme