AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

CSG: Cognitive Structure Generation for Intelligent Education

Hengnian Gu, Zhifu Chen, Yuxin Chen, Jin Peng Zhou, Dongdai Zhou

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Probabilistic embeddings and relational inference

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 pretraining on simulated structures to the generic-reward RL variant (ablation)CSDPM (V3 to V5) · NIPS34 KTAUC: 0.7196 -> 0.7318Improved
add pretraining on simulated structures to the generic-reward RL variant (ablation)CSDPM (V3 to V5) · NIPS34 CDAUC: 0.7857 -> 0.7997Improved
add generic-reward RL to the pretrained model (ablation)CSDPM (V2 to V5) · NIPS34 KTAUC: 0.6951 -> 0.7318Improved
add pretraining to the learning-free simulated structures (ablation)CSDPM (V1 to V2) · NIPS34 KTAUC: 0.6807 -> 0.6951Improved
replace generic reward with SOLO hierarchical reward, no pretraining (ablation)CSDPM (V3 to V4) · NIPS34 KTAUC: 0.7196 -> 0.7242Improved
replace generic reward with SOLO hierarchical reward, with pretraining (ablation)CSG (V5 to CSG) · NIPS34 KTAUC: 0.7318 -> 0.7413Improved
replace generic reward with SOLO hierarchical reward, with pretraining (ablation)CSG (V5 to CSG) · NIPS34 CDAUC: 0.7997 -> 0.8036Improved
replace generic reward with SOLO hierarchical reward, with pretraining (ablation)CSG (V5 to CSG) · ASSISTment17 CDAUC: 0.7968 -> 0.8003Improved
use CSG structures instead of GPT-5 generated graphsdownstream KT · FrcSub (500 students)AUC: 0.7430 -> 0.8602Improved
use CSG structures instead of Llama-3-70B generated graphsdownstream KT · FrcSub (500 students)AUC: 0.5170 -> 0.8602Improved
add Stage II SOLO RL to simulated structures (ablation)CSG structure quality · FrcSubJaccard vs human annotation: 0.46 -> 0.79Improved
use CSG instead of GPT-5 for structure generationstructure quality · Math2Graph edit distance vs human annotation: 0.43 -> 0.21Improved

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