COMPASS: Context-Modulated PID Attention Steering System for Hallucination Mitigation

Kenji Sahay, Snigdha Pandya, Rohan Nagale, Anna Lin, Shikhar Shiromani, Kevin Zhu, Sunishchal Dev

Keywords: Large Language Models (LLMs), Contextual Hallucination, Attention Modulation, PID Control Feedback Loop, Factual Accuracy, Real-Time Intervention, Cross-Model Evaluation, Cross-Dataset Generalization, LLaMA, Mistral, Qwen

TL;DR: We steer LLMs away from hallucinations by dynamically amplifying context-sensitive attention heads.

Abstract:

Large language models (LLMs) often generate fluent
but factually incorrect statements despite having access
to relevant evidence, a failure mode rooted in how they
allocate attention between contextual and parametric
knowledge. Understanding and steering this internal
behavior is key both for trustworthy deployment and
for scientific interpretability of model mechanisms.
We introduce COMPASS (Context-Modulated PID At-
tention Steering System), a lightweight, interpretable
control framework that embeds a model-based feedback
loop directly within decoding. COMPASS quantifies
context reliance via a transparent metric, the Context
Reliance Score (CRS), which serves as an online probe
of how attention heads ground generation in evidence.
Using this interpretable signal, a PID controller dynam-
ically modulates attention heads to maintain factual
consistency without retraining or multi-pass decoding.
Across benchmarks (HotpotQA, XSum, HaluEval,
RAGTruth), COMPASS consistently reduces contextual
hallucination rates (2.8–5.8% absolute) while revealing
how distinct attention heads contribute to evidence
alignment. These results highlight feedback-driven in-
terpretability as a pathway toward scientific understand-
ing of LLM behavior.

Submission Number: 15