AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

VeriRole: Verifiable Role-Awareness through Hint-Guided Reinforcement Learning

Zongsheng Wang, Kaili Sun, Bowen Wu, qun yu, Ying Li, Xu Chen, Baoxun Wang

ICLR 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards

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
add hint reward (ablation)Qwen2.5-32B-GRPO · RAIDENaverage score: 0.6598 -> 0.8268Improved
add hint reward (ablation)Qwen2.5-32B-GRPO · RAIDEN Topic Advancementscore: 0.2089 -> 0.6865Improved
add accuracy reward (ablation)Qwen2.5-32B-GRPO · RAIDENaverage score: 0.8061 -> 0.8268Improved
train on RAIDEN plus Situation Puzzle instead of Situation Puzzle only (ablation)Qwen2.5-32B-GRPO · RAIDENaverage score: 0.7111 -> 0.8268Improved
add Situation Puzzle data to RAIDEN-only training (ablation)Qwen2.5-32B-GRPO · RAIDENaverage score: 0.8143 -> 0.8268Improved
use GRPO instead of SFT on repliesQwen2.5-32B · RAIDENaverage score: 0.6809 -> 0.8268Improved
use GRPO instead of SFT on repliesQwen2.5-32B · RAIDEN Topic Advancementscore: 0.2835 -> 0.6865Improved
apply GRPO to instruct modelQwen2.5-14B · RAIDENaverage score: 0.6302 -> 0.7725Improved
apply GRPO to role-play modelPeach-9B-roleplay · RAIDENaverage score: 0.3611 -> 0.6183Improved
apply GRPO to a model with inherent reasoningQwen3-32B · RAIDEN Topic AdvancementscoreHurt
enable thinking modeQwen3-32B · RAIDENaverage score: 0.6989 -> 0.6796No clear effect
GRPO training before SFT on psychotherapist dataQwen2.5-32B · RAIDEN after post-training SFTaverage score: 0.6366 -> 0.8006Improved

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