AI Paper Atlas

Research map / Large language models / Alignment and preferences

Aligning Language Models with Human Preferences via a Bayesian Approach

Jiashuo WANG, Haozhao Wang, Shichao Sun, Wenjie Li

NeurIPS 2023 · Large language models › Alignment and preferences · approach: Reward modeling for diverse preferences · 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 d-PM instead of majority-vote preference model (ablation)Aligned MultiESC · ESConvBLEU-4: 2.88 -> 3.36Improved
use d-PM instead of soft-label preference model (ablation)Aligned MultiESC · ESConvCIDEr: 30.75 -> 32.65Improved
use d-PM instead of unaggregated annotations (w/oA) (ablation)Aligned MultiESC · ESConvCIDEr: 28.58 -> 32.65Improved
align MultiESC with majority-vote preference model (ablation)MultiESC · ESConvBLEU-4: 3.14 -> 2.88Hurt
align MultiESC with d-PM calibrationMultiESC · ESConv human evalGlobal Consensus: 2.693 -> 2.783Improved
align Blender-Vanilla with d-PMBlender-Vanilla · ESConvExtreme: 51.02 -> 50.38Hurt
use majority-vote model for Blender-Joint (ablation)Blender-Joint · ESConvCIDEr: 16.91 -> 21.86Improved
use w/oA model for Blender-Joint (ablation)Blender-Joint · ESConvCIDEr: 16.91 -> 15.29Hurt
align T5-small with d-PM (greedy)T5-small · MICROUGE-1: 37.04 -> 38.15Improved
align BART-large with soft model (beam) (ablation)BART-large · MICROUGE-2: 35.07 -> 34.85Hurt
align Flan-T5 with d-PM (beam)Flan-T5-base · MICROUGE-1: 55.07 -> 55.18Improved
align BART-large with d-PM (beam)BART-large · MICAvg.Len: 9.44 -> 9.40No 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