Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators
This addresses fairness and reliability issues in automated evaluation pipelines for tasks like preference tuning and model routing, but it is incremental as it builds on existing steering methods.
The paper tackled the problem of self-preference bias in LLM evaluators, where models favor their own outputs, and found that lightweight steering vectors at inference time can reduce unjustified bias by up to 97%.
Large language models (LLMs) increasingly serve as automated evaluators, yet they suffer from "self-preference bias": a tendency to favor their own outputs over those of other models. This bias undermines fairness and reliability in evaluation pipelines, particularly for tasks like preference tuning and model routing. We investigate whether lightweight steering vectors can mitigate this problem at inference time without retraining. We introduce a curated dataset that distinguishes self-preference bias into justified examples of self-preference and unjustified examples of self-preference, and we construct steering vectors using two methods: Contrastive Activation Addition (CAA) and an optimization-based approach. Our results show that steering vectors can reduce unjustified self-preference bias by up to 97\%, substantially outperforming prompting and direct preference optimization baselines. Yet steering vectors are unstable on legitimate self-preference and unbiased agreement, implying self-preference spans multiple or nonlinear directions. This underscores both their promise and limits as safeguards for LLM-as-judges and motivates more robust interventions.