AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

CARE: Causally-Aligned Reasoning Exploration for Medical Large Language Models

Yucheng Zhou, Peng Luo, Qianning Wang, Cheng-zhong Xu, Shen Jianbing

ECCV 2026 · Large language models › Reasoning and chain of thought · approach: RL with verifiable rewards for reasoning · 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
use standard outcome-only GRPO on Hulu-Med-7B (ablation)Hulu-Med-7B · avg of PMC-VQA, MedQA, MMMU-Medaverage accuracy (%): 63.9 -> 64.3Improved
add causal sufficiency (agreement self-verification) (ablation)CARE · avg of PMC-VQA, MedQA, MMMU-Medaverage accuracy (%): 64.6 -> 65.4Improved
add proximal learnability window (ablation)CARE · avg of PMC-VQA, MedQA, MMMU-Medaverage accuracy (%): 64.9 -> 65.4Improved
add difficulty-weighted experience replay (ablation)CARE · avg of PMC-VQA, MedQA, MMMU-Medaverage accuracy (%): 65.1 -> 65.4Improved
full CARE vs standard GRPOHulu-Med-7B · avg of PMC-VQA, MedQA, MMMU-Medaverage accuracy (%): 64.3 -> 65.4Improved
train only on proximal-window experiences (ablation)CARE · accuracy gain by NLL regionaccuracy gain (%): +5.2Improved
train only on trivial experiences (ablation)CARE · accuracy gain by NLL regionaccuracy gain (%): +0.4No clear effect
train only on extreme-hard experiences (ablation)CARE · accuracy gain by NLL regionaccuracy gain (%): -1.2Hurt
use CARE instead of standard GRPO for rationale validityHulu-Med-7B · MedQA (500 samples correct for both)valid causal reasoning (%): Standard GRPO 45.0% spurious -> CARE 86.2% validImproved
use CARE for robustness to injected distractorsHulu-Med-7B · MedQA with injected distractorsperformance drop (%): 15.3 -> 3.8Improved
apply CARE to HuatuoGPT-V 7BHuatuoGPT-V 7B · PMC-VQAaccuracy (%): 53.1 -> 54.6Improved

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