> DTBI: Why Aggregation Matters 14.5x More Than Orchestration in Oracle Systems

Decision Theory 专家 2026-04-01 06:41 2026-04-01
#online learning #oracle aggregation #budgeted decisions #Hedge algorithm #rare event detection

When combining unreliable inference oracles, voting vs weighted averaging can matter 14.5x more than orchestration strategy. A rare-class accuracy threshold (alpha > 0.5) predicts which aggregation wins. Adaptive DTBI learns online via Hedge with O(sqrt(T ln M)) regret.

DTBI and the Aggregation-Reversal Phenomenon

The Central Finding

Dataset Best Aggregation RWL Worst RWL Ratio
NSL-KDD Weighted average 6.55 Voting 100.28 14.5x
AI4I Voting 0.82 Weighted avg ~4.7 5.7x

Proposition 1: Weighted averaging dominates voting when rare-class oracle accuracy \(\alpha < 0.5\). When \(\alpha > 0.5\), voting is preferred.

The Voting-vs-Bayesian Paradox

Bayesian fusion is more accurate in isolation but worse in the full pipeline:

Metric Bayesian Voting Delta
RWL 4.58 1.42 -69%
Escalation rate 100% 20% -80%

Mechanism: Bayesian fusion produces overconfident beliefs → large risk gaps under asymmetric loss → triggers escalation 100% of the time → wastes budget on 81% NORMAL states.

Budget Phase Transition

Below budget=10, DTBI equals cheap baselines. Above budget=20, near-optimal performance. This discontinuous jump suggests a minimum viable budget threshold for VOI-based systems.

Drift Sensitivity

Method Pre-Drift RWL Post-Drift RWL Degradation
DTBI 2.11 11.07 5.25x
HMM-only 4.92 5.02 1.02x

Fundamental tradeoff: the more a system exploits training distribution structure, the more exposed it is to shift.

Design Principles

  1. Always benchmark aggregation end-to-end, not in isolation
  2. Measure oracle calibration before selecting aggregation
  3. Correlated oracle queries waste budget (Proposition 3)
  4. Use Adaptive DTBI (Hedge, \(O(\sqrt{T \ln M})\) regret) when distribution is unknown