EPC: A Standardized Protocol for Measuring Evaluator Preference Dynamics in LLM Agent Systems
2026-07-01 • Machine Learning
Machine LearningComputation and Language
AI summaryⓘ
The authors study how biases in feedback systems used by AI language models can affect the models' behavior over time, a problem called evaluator preference coupling. They created a clear and detailed procedure (a protocol) to help researchers measure and compare these biases consistently, even as the tools change. They also provide a snapshot of real measurements from different models and evaluators, along with guidelines and code, so others can reproduce and track these effects. This work aims to make it easier to understand and detect changes in evaluator bias over time.
LLM agentsevaluator feedbackevaluator preference couplingclosed-loop adaptationprotocol specificationTTRL update rulemetric computationmeasurement decayAPI endpointsversioning convention
Authors
Zewen Liu
Abstract
When LLM agents use evaluator feedback to adapt their behavior in closed loops, evaluator biases propagate through the agent's strategy distribution -- a phenomenon known as evaluator preference coupling. Prior work has documented coupling across multiple evaluator families and model versions, but the field lacks a standardized protocol that enables third-party researchers to (i) reproduce coupling measurements, (ii) compare results across evaluators and time points, and (iii) detect measurement decay as proprietary evaluators silently update. This paper provides the protocol. We specify EPC (Evaluator Preference Coupling) -- a detailed, RFC-style protocol specification for the four-phase isolation paradigm, covering executor and evaluator configuration, strategy and task design, the TTRL update rule, metric computation (gamma, JSD, ECE, Brier), and output schema. We accompany the protocol with a versioned Reference Snapshot v1.0: coupling measurements for eight evaluator conditions (N=122 unique experimental repetitions across GPT-4o, Qwen, DeepSeek, and others) derived from five independent studies, annotated with evaluator version identifiers, API endpoints, and measurement dates. The snapshot is explicitly time-bound: all values are conditional on specific model versions and are expected to decay as proprietary evaluators update. We define a versioning convention (vX.Y-Z, encoding protocol version, snapshot version, and evaluator generation) and provide a usage guide covering adoption, interpretation, and known pitfalls. The protocol, reference snapshot, and implementation code are released as open infrastructure.