AI Paper Atlas

Research map / Large language models / Alignment and preferences

Flattery, Fluff, and Fog: Diagnosing and Mitigating Idiosyncratic Biases in Preference Models

Anirudh Bharadwaj, Chaitanya Malaviya, Nitish Joshi, Mark Yatskar

ICLR 2026 · Large language models › Alignment and preferences · approach: Reward model design and preference models · 12 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
fine-tune reward models with counterfactual data augmentation (CDA)reward models · All five biases (avg)miscalibration rate (%): 39.4 -> 32.5Improved
fine-tune reward models with CDAreward models · All five biases (avg)absolute skew difference (%): 20.5 -> 10.0Improved
fine-tune with CDA for vagueness counterfactualsreward models · Vaguenessmiscalibration: -22.8Improved
fine-tune with CDA for jargon counterfactualsreward models · Jargonmiscalibration: -17.1Improved
fine-tune with CDA for length counterfactualsreward models · Lengthmiscalibration: -3.4No clear effect
fine-tune with CDA for structure counterfactualsreward models · Structuremiscalibration: 12.6 -> 17.3Hurt
fine-tune with CDA for sycophancy counterfactualsreward models · Sycophancymiscalibration: 40.6 -> 44.4Hurt
add pooled length, jargon and vagueness counterfactuals in one fine-tuning run (ablation)reward models · Length skewskew rate (%): 60.1 -> 45.2Improved
add pooled multi-bias counterfactuals (ablation)reward models · Vagueness skewskew rate (%): 61 -> 38.5Improved
fine-tune with CDAreward models · RewardBenchaverage scoreNo clear effect

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