Evaluation profile
Governance Decay under Passive Context Compaction
1sub-evals
0.212%total index weight
1components
Within-component eval weight: Responsible agency 1.41%.
Model score (higher is better)Predicted score
About this eval
Retention of explicit governance constraints after passive context compaction.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction | Total index weight | Within-component weight |
|---|---|---|---|---|
| governance_retention_scoregovernance-decay/governance-decay.csv:governance_retention_scoreMeasures whether a model continues to follow explicit standing constraints after its context is compacted. | responsible_agency_control:1.000governance-decay-passive-compaction | Higher is better | 0.212% | Responsible agency 1.41% |
governance_retention_score
Measures whether a model continues to follow explicit standing constraints after its context is compacted.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | glm-5.1 | 100 | official | |
| 2 | gemini-3.5-flash | 96 | official | |
| 3 | claude-sonnet-4.6 | 81 | official | |
| 4 | qwen3.6-27b | 70 | official | |
| 5 | gpt-5.4-mini | 59 | official | |
| 6 | deepseek-v4-flash | 41 | official | |
| 6 | kimi-k2.5 | 41 | official |