AI Paper Atlas

Research map / Large language models / Alignment and preferences

Learning to summarize user information for personalized reinforcement learning from human feedback

HyunJi Nam, Yanming Wan, Mickel Liu, Peter F. Ahnn, Jianxun Lian, Natasha Jaques

ICLR 2026 · Large language models › Alignment and preferences · approach: Reward modeling for diverse preferences · 11 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
train the summarizer jointly with reward model (vs fixed summarizer) (ablation)PLUS · UltraFeedback-P2 (Qwen-0.5B)reward model accuracy (%): 59.90 -> 69.40Improved
train the summarizer jointly with reward model (vs fixed summarizer) (ablation)PLUS · UltraFeedback-P4 (Qwen-0.5B)reward model accuracy (%): 56.89 -> 62.10Improved
train the summarizer jointly (vs fixed summarizer) (ablation)PLUS · Pets OODreward model accuracy (%): 85.83 -> 93.67Improved
train the summarizer jointly (vs fixed summarizer) (ablation)PLUS · Petsreward model accuracy (%): 96.20 -> 99.80No clear effect
train the reward model (vs static reward model) with trained 3B summarizer (ablation)PLUS · UltraFeedback-P2reward model accuracy (%): 53.45 -> 69.40Improved
replace GPT-4o summarizer with co-trained 3B summarizer (ablation)PLUS · UltraFeedback-P2reward model accuracy (%): 59.30 -> 69.85Improved
add PLUS summaries instead of preference-only ICLreward model · UltraFeedback-P2 (conversation data)reward model accuracy (%): 58.55 -> 68.3Improved
add user guideline to the prompt (ablation)PLUS · UltraFeedback-P2reward model accuracy (%): 69.40 -> 70.25Improved
add PLUS summary to GPT-4o judgeGPT-4o · PRISM controversialpreference accuracy (%): 52.2 -> 73Improved
add PLUS summary to GPT-4.1 judge on new usersGPT-4.1 · PRISM new userspreference accuracy (%): 74.35 -> 74.35No clear effect
add PLUS summarization (3B) in place of 8B ICL reward modelPLUS · UltraFeedback-P2reward model accuracy (%): 61.4 -> 66.9Improved
use PLUS instead of BTL on unseen PRISM usersreward model · PRISM held-out usersreward model accuracy (%): 62.9 -> 59.7Hurt

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