AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs

Lecheng Yan, Ruizhe Li, Guanhua Chen, Qing Li, Jiahui Geng, Wenxi Li, Longyue Wang, Chenyang Lyu

ICML 2026 · Large language models › Language, safety and interpretability · approach: Reasoning mechanisms and failure modes · 0 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
keep L18-20 MLP weights from RLVR (vs resetting Anchor to base) (ablation)spurious-RLVR Qwen2.5-Math-7B · MATH-500 leakage subsetaccuracy: 98% -> 86% when resetImproved
keep L21-22 MLP weights from RLVR (vs resetting Adapter to base) (ablation)spurious-RLVR Qwen2.5-Math-7B · leakage subsets MATH-500/MinervaMathaccuracy: drops of -6% and -10% when resetImproved
keep only Anchor or only Adapter RLVR layers (ablation)spurious-RLVR Qwen2.5-Math-7B · MinervaMathaccuracy: Keep Only Adapter 32%, Keep Only Anchor 42%, Keep Both 46% (base 50%)Hurt
reset Anchor/Adapter layers (ablation)spurious-RLVR Qwen2.5-Math-7B · LiveMathBench (leakage-free)accuracy: stable at 70%No clear effect
reset Anchor layers (ablation)spurious-RLVR Qwen2.5-Math-7B · stable (non-leaked) samplesaccuracyNo clear effect
reset Anchor layers (ablation)spurious-RLVR Qwen2.5-Math-1.5B · MATH-500 leakage subset (38 cases)accuracy: 39.5% -> 2.6% when resetImproved
scale key activations of top-10 task neurons at Layer 18 (suppress/amplify)spurious-RLVR Qwen2.5-Math-7B · leakage datasetsaccuracy: -3.8% suppression, +4.4% amplificationMixed
scale neuron keys at Layer 18spurious-RLVR Qwen2.5-Math-7B · leakage-free datasetaccuracyNo clear effect
scale neuron keys at Layer 25 (Structural Adapter)spurious-RLVR Qwen2.5-Math-7B · leakage datasetsaccuracyHurt
add probe-gated Layer-19 neuron suppressionRLVR-tuned Qwen2.5-Math-7B · MATH-500 wrong->rightaccuracy: 100.0% -> 85.16%Hurt
add probe-gated Layer-19 neuron suppressionRLVR-tuned Qwen2.5-Math-7B · AIME-2025 / LiveMathBenchaccuracy: 30.0% -> 30.0%; 21.0% -> 21.0%No 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