AI Paper Atlas

Research map / Large language models / Alignment and preferences

The Alignment Auditor: A Bayesian Framework for Verifying and Refining LLM Objectives

Matthieu Bou, Nyal Patel, Arjun Jagota, Satyapriya Krishna, Sonali Parbhoo

ICLR 2026 · Large language models › Alignment and preferences · approach: Steering and correcting LLM behavior · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use sequential Bayesian updates instead of single round (ablation)Alignment Auditor (Llama-3.2-1B) · toxicity auditposterior log-det covariance: -196 -> -897Improved
use sequential Bayesian updates instead of single round (ablation)Alignment Auditor (Llama-3.2-1B) · toxicity auditepistemic uncertainty (MI): 0.1272 -> 0.0683Improved
use sequential Bayesian updates instead of single round (ablation)Alignment Auditor (Llama-3.2-1B) · toxicity auditsingle-text AUROC: 0.7954 -> 0.8323Improved
use sequential Bayesian updates instead of single round (ablation)Alignment Auditor (Llama-3.2-1B) · toxicity auditpairwise accuracy: 0.7597 -> 0.7524No clear effect
scale base model from Llama-3.2-1B to Llama-3.1-8BAlignment Auditor · toxicity auditsingle-text AUROC: 0.832 -> 0.916Improved
scale base model from Llama-3.2-1B to Llama-3.1-8BAlignment Auditor · toxicity auditCohen's d: 1.325 -> 1.821Improved
scale base model from Llama-3.2-1B to Llama-3.1-8BAlignment Auditor · toxicity auditpairwise ECE: 0.0425 -> 0.0462Hurt
scale base model from Llama-3.2-1B to Llama-3.1-8BAlignment Auditor · HH-RLHF helpfulnesspairwise accuracy: 0.725 -> 0.729Improved
scale base model from Llama-3.2-1B to Llama-3.1-8BAlignment Auditor · HH-RLHF helpfulnesspairwise ECE: 0.0328 -> 0.0438Hurt
use full-posterior reward variance instead of output toxicity probability for spurious-marker detectionAlignment Auditor · Spurious-marker detectionAUC: 0.58 -> 0.88Improved
use posterior-mean reward instead of output toxicity probabilityAlignment Auditor · Spurious-marker detectionAUC: 0.58 -> 0.70Improved
use round 1 posterior instead of contracted posterior (rounds 2-5) as RLHF reward (ablation)Alignment Auditor · toxicity PPO alignmentreward hackingHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy