AI Paper Atlas

Research map / Large language models / Alignment and preferences

Distributionally Robust Reinforcement Learning from Human Feedback

Debmalya Mandal, Paulius Sasnauskas, Goran Radanovic

ICML 2026 · Large language models › Alignment and preferences · approach: Active and token-level preference learning

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 reward model with distributionally robust loss (rho=0.2)Bradley-Terry reward model · RewardBench (reward model, 400K Unified-Feedback)Avg accuracy: 72.3 -> 72.9Improved
train reward model with distributionally robust loss (rho=0.2)Bradley-Terry reward model · RewardBench (reward model, 400K Unified-Feedback)Reasoning accuracy: 66.8 -> 71.4Improved
increase robustness radius rho to 1.0 for reward model (ablation)DR reward model · RewardBench (reward model, 400K Unified-Feedback)Avg accuracy: 72.3 -> 46.4Hurt
increase robustness radius rho from 0.2 to 0.4 for reward model (ablation)DR reward model · RewardBench (reward model, 400K Unified-Feedback)Avg accuracy: 72.9 -> 61.0Hurt
use DR PPO with rho=0.1 instead of standard PPOPPO with BTL reward model · RewardBench (PPO policy, 5K Unified-Feedback)Reasoning accuracy: 45.9 -> 55.9Improved
use DR PPO with rho=0.1 instead of standard PPOPPO with BTL reward model · RewardBench (PPO policy, 5K Unified-Feedback)Avg accuracy: 48.7 -> 52.3Improved
use DR PPO rho=0.6 instead of standard PPO (ablation)PPO · RewardBench (PPO policy, 5K Unified-Feedback)Avg accuracy: 48.7 -> 48.8No clear effect
use DR PPO rho=0.1 instead of standard PPOPPO · HHH Alignment (PPO)accuracy: 66.9 ± 0.5 -> 68.1 ± 0.3Improved
use DR PPO rho=0.1 instead of standard PPOPPO · MT Bench (PPO)accuracy: 59.9 ± 0.3 -> 61.6 ± 1.0Improved
use DR PPO rho=0.1 instead of standard PPOPPO · Unified-Feedback test (PPO)accuracy: 59.8 ± 0.2 -> 59.8 ± 0.4No clear effect
use DR DPO rho=0.8 instead of standard DPODPO · Unified-Feedback test (DPO)accuracy: 60.9 ± 0.02 -> 64.4 ± 0.3Improved
use DR DPO rho=0.8 instead of standard DPODPO · HHH Alignment (DPO)accuracy: 69.2 ± 0.1 -> 70.6 ± 0.4Improved

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