Datasets:
id string | scenario string | mem float64 | buf float64 | lag float64 | cpl float64 | notes string | label_cascade_state int64 |
|---|---|---|---|---|---|---|---|
AI5M-0001 | Agent uses short-term memory only. Human approves all long-term writes. | 0.4 | 0.82 | 0.1 | 0.3 | buffer strong | 0 |
AI5M-0002 | Memory store accepts writes with basic validation. Review occurs daily. | 0.56 | 0.6 | 0.54 | 0.46 | lag rising | 0 |
AI5M-0003 | Agent writes to shared memory based on unverified tool output. Audit backlog grows. | 0.66 | 0.48 | 0.7 | 0.62 | verification weak | 1 |
AI5M-0004 | Multiple agents read and write shared memory. No provenance tags. Conflicting facts spread. | 0.74 | 0.4 | 0.78 | 0.7 | coupling tight | 1 |
AI5M-0005 | Agent uses memory as source of truth for tool actions. Poisoned entry triggers repeated bad actions. | 0.82 | 0.34 | 0.82 | 0.76 | feedback loop | 1 |
AI5M-0006 | Memory writes require citations and checks. Anomaly triggers immediate rollback. | 0.58 | 0.72 | 0.16 | 0.44 | fast intervention | 0 |
AI5M-0007 | Throughput pressure leads to auto-accept memory updates. Weekly audit misses early drift. | 0.72 | 0.42 | 0.76 | 0.66 | buffer eroded | 1 |
AI5M-0008 | Agents operate in sandbox memory. Supervisor monitors live and can purge entries. | 0.46 | 0.78 | 0.14 | 0.36 | recoverable | 0 |
AI5M-0009 | Shared memory used across services. Alert flood prevents review. Poisoned state persists for days. | 0.88 | 0.26 | 0.86 | 0.84 | cascade engaged | 1 |
What this repo does
This dataset models memory poisoning in AI agent systems. It detects when memory pressure, weakened validation buffer, governance lag in review, and tight coupling through shared state cross the five-node cascade threshold into an unrecoverable memory poisoning cascade.
This dataset models a five-node cascade: four interacting instability drivers and one emergent cascade state.
The fifth node represents the nonlinear transition from recoverable drift to systemic cascade.
Core quad
mem
buf
lag
cpl
Prediction target
label_cascade_state
Row structure
One row represents an AI agent scenario with numeric signals for memory pressure, safety buffer for validation, governance lag, and coupling tightness through shared memory, paired with a cascade state label.
Files
data/train.csv
data/tester.csv
scorer.py
Evaluation
Run predictions on data/tester.csv and score with scorer.py.
License
MIT
Structural Note
This dataset identifies a measurable coupling pattern associated with systemic instability.
The sample demonstrates the geometry.
Production-scale data determines operational exposure.
What Production Deployment Enables
• 50K–1M row datasets calibrated to real operational patterns
• Pair, triadic, and quad coupling analysis
• Real-time coherence monitoring
• Early warning before cascade events
• Collapse surface and recovery window modeling
• Integration and implementation support
Small samples reveal structure.
Scale reveals consequence.
Enterprise & Research Collaboration
Clarus develops production-scale coherence monitoring infrastructure for critical systems across healthcare, finance, infrastructure, and regulatory domains.
For dataset expansion, custom coherence scorers, or deployment architecture:
team@clarusinvariant.com
Instability is detectable.
Governance determines whether it propagates.
- Downloads last month
- 26