AI Paper Atlas

Research map / Large language models / Alignment and preferences

Mitigating Length Bias in RLHF Through a Causal Lens

Hyeonji Kim, Sujeong Oh, Sanghack Lee

AAAI 2026 · Large language models › Alignment and preferences · approach: Reward modeling for diverse preferences · 4 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 model on counterfactual data (full fine-tune of HRO)HRO reward model · Chatbot Arena length-controlledLC accuracy: 0.249 -> 0.493Improved
fine-tune reward model on counterfactual dataHRO reward model · RewardBench-1average accuracy: 0.486 -> 0.506Improved
use LoRA instead of full fine-tuning on counterfactual data (ablation)CDA reward model · Chatbot Arena length-controlledLC accuracy: 0.493 -> 0.248Hurt
train policy with CDA_HRO reward model instead of HROOpenLLaMA-3B PPO · AlpacaEvallength-controlled winrate (%): 18.97 -> 37.18Improved
train policy with CDA_HRO reward model instead of HROOpenLLaMA-3B PPO · AlpacaEvalaverage output length: 2048 -> 1118Improved
use CDA reward instead of ODIN for PPOOpenLLaMA-3B PPO · AlpacaEvalwinrate (%): 11.34 -> 32.55Improved
train reward model on counterfactual datareward model · RewardBench-2accuracyNo 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