AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Detecting Data Contamination from Reinforcement Learning Post-training for Large Language Models

Yongding Tao, Tian Wang, Yihong Dong, Huanyu Liu, Kechi Zhang, Hu XiaoLong, Ge Li

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 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
use self-critique probing with entropy similarity instead of best baselinecontamination detectors · RL-MIA (Qwen2.5-7B-Instruct)average AUC: 0.62 -> 0.70Improved
use self-critique probing with entropy similarity instead of best baselinecontamination detectors · RL-MIA (DeepSeek-Math-7B-Instruct)average AUC: 0.55 -> 0.64Improved
likelihood-based detectors (PPL, Min-K%, Min-K%++) for RL-phase contaminationdetectors · RL-MIAAUC: 0.44 to 0.53 average AUCNo clear effect
replace CDD sampling probe with entropy (Entropy-Temp) (ablation)CDD · RL-MIA (Qwen2.5-7B-Instruct)average AUC: 0.51 -> 0.56Improved
replace Recall metric with entropy (Entropy-Noise) (ablation)Recall · RL-MIA (Qwen2.5-7B-Instruct)average AUC: 0.59 -> 0.62Improved
replace Recall metric with entropy (Entropy-Noise) (ablation)Recall · RL-MIA (DeepSeek-Math-7B-Instruct)average AUC: 0.54 -> 0.55Improved
use self-critique probing instead of non-member prefix or temperature sampling for entropy detectors (ablation)Entropy-Temp/Entropy-Noise · RL-MIAaverage AUC: +13% (Qwen2.5-7B-Instruct), +16% (DeepSeek-Math-7B-Instruct)Improved
train with different RL algorithms (PPO, GRPO, DAPO) (ablation)Self-Critique · K&K (Qwen2.5-3B-Instruct)AUC: 0.61, 0.61, 0.60No clear effect
reduce Top-K entropy approximation from K=50 to K=3 (ablation)Self-Critique · AIME25 (Qwen2.5-7B-Instruct)AUC: 0.7156 -> 0.7022No clear effect
reduce Top-K entropy approximation from K=50 to K=3 (ablation)Self-Critique · K&K (Qwen2.5-7B-Instruct)AUC: 0.6584 -> 0.6460No clear effect
restrict to lower-pretraining-contamination subset (ablation)Self-Critique · GSM8K (Qwen2.5-0.5B-Instruct, PPO)AUCImproved
random-control subset of same size (ablation)Self-Critique · GSM8K (Qwen2.5-0.5B-Instruct, PPO)AUCNo 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