AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Cold-Start Personalization via Bayesian Adaptive Questioning

Avinandan Bose, Shuyue Stella Li, Faeze Brahman, Pang Wei Koh, Simon Shaolei Du, Yulia Tsvetkov, Maryam Fazel, Lin Xiao, Asli Celikyilmaz

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Cold-start recommendation models

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
learn offline world model of preference correlations (latent structure) instead of none (ablation)CAPE · PrefDiscogap closed to ground truthImproved
use adaptive querying instead of random querying (ablation)CAPE · PrefDiscogap closed to ground truth at T=5: 24% vs 21% at T=5Improved
use adaptive question selection (ablation)CAPE · PrefDisconumber of questions needed: saves one question per interaction (T=5 matches random at T=6)Improved
replace end-to-end GRPO RL questioning policy with offline world model plus Bayesian inferencepreference elicitation · PrefDisco average over four domainspreference alignment: 68.5% -> 80.8%Improved
replace GRPO with CAPEpreference elicitation · CSQApreference alignment (% of Oracle): 55.2 -> 78.2Improved
replace GRPO with CAPEpreference elicitation · query efficiencyqueries for equal alignment: 2.5x on AIME, 7x on MedQA, >15x on SocialIQA and CSQAImproved
GRPO end-to-end RL training vs population averageLlama-3.1-8B-Instruct · SocialIQA, CSQApreference alignment (% of Oracle): SocialIQA 82.1 -> 71.3; CSQA 72.4 -> 55.2Hurt
CollabLLM generic interactivity trainingLlama-3.1-8B-Instruct · PrefDiscopreference alignment (% of Oracle): 20-26% alignment despite 28-75% adaptivityHurt
use CAPE instead of GRPO on adaptivityquestion selection · PrefDiscoadaptivity (%): GRPO 0-28% vs CAPE 39-62%Improved

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