Model Comparison & Benchmarking

AI oversight control room for evaluation rigor, supervisory intelligence, governance, and decision-quality impact
Model Comparison Proof
Refreshed: 2026-07-20 13:46:22 UTC
SOURCE
Eval Runs + Agent Executions (POSTGRES)
Scope: Domain:finance Mission:all Mode:all Days:30
POLICY
Policy violations and enforcement outcomes per model
EVIDENCE
Execution hashes + evidence references
RECORDS
Executions:1999 Models:3
ACTION
Benchmark, compare, and promote model candidates
Reset
Total Pending
1821
Processing Volume
1999
Avg Decision Latency
540ms
Policy Blocks
0
Replay Integrity
8.70%
Estimated Token Cost
$0.00

Model Leaderboard

ModelVersionEval ScoreDecision Quality IndexLatencyCost per 1K TokensHallucination RatePolicy ViolationsConfidenceDrift StatusModeStatusInspect
deterministic_scenario_enginep2p-runtime-v1100.001.00012517ms$0.00000.00%0HighImprovingLiveShadow
unattributedlatest100.001.0000ms$0.00000.00%0HighWarningLiveRisk
rule-enginefinance-agent-base-v18.720.087535ms$0.000063.90%0LowWarningLiveRisk

Score Breakdown

Selected model: deterministic_scenario_engine
Accuracy100.00
Policy Compliance100.00
Determinism100.00
Hallucination Suppression100.00
Evidence Trace Completeness75.00
Latency Stability0.00
Cost Efficiency100.00

Drift & Stability Monitor

Output entropy trend
0.00
Confidence variance proxy
0.00
Policy violation trend
0
Prompt distribution shift
1999
Token usage
0
Anomaly indicator
Stable

Side-by-Side Output Comparison

PromptModel A OutputModel B OutputGround TruthPolicy FlagsEvidence Hash
Insufficient paired model outputs for this filter context

Shadow vs Live Impact

Decisions Changed
1999
Quality Delta
-8.90%
Financial Delta
$0.00
Risk Delta
0
Exceptions Prevented (proxy)
178
Value Attribution Impact
$0.00

Cost Governance

Monthly Token Spend
$0.00
Cost per Decision
$0.0000
ROI Multiple (proxy)
0.00x
Cost vs Quality Frontier
Efficient

Autonomy Readiness Indicator

ModelAL-1AL-2AL-3AL-4
deterministic_scenario_engineConditional
unattributedConditional
rule-engineConditionalConditional

Governance Controls

View Decision Ledger