Research map / Label-efficient and robust learning / Self-supervised and contrastive
Invariant Representation Learning for Memory Behavior Modeling via Adaptive Environment Separation
Xiaoxuan Shen, Zhihai Hu, Fuqing Li, Shengyingjie Liu, Jianwen Sun
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add the InfoMax encoder stage (vs random encoder) (ablation) | I-Mem · Anki | AUC: 10% drop in AUC | Improved |
| add the mapping step (vs using high-dimensional representations directly) (ablation) | I-Mem · Anki | AUC: 24% decrease | Improved |
| increase spurious set size from H-1 to H-3 (ablation) | I-Mem · Anki | AUC: 0.7037 -> 0.7139 | Improved |
| increase spurious set size from H-1 to H-2 (ablation) | I-Mem · MaiMemo | AUC: 0.5726 -> 0.5969 | Improved |
| increase spurious set size from H-2 to H-5 (ablation) | I-Mem · MaiMemo | AUC: 0.5969 -> 0.5692 | Hurt |
| increase spurious set size from H-1 to H-5 (ablation) | I-Mem · En2Es | AUC: 0.6918 -> 0.6812 | Hurt |
| increase spurious set size from H-1 to H-5 (ablation) | I-Mem · En2Fr | AUC: 0.6691 -> 0.6691 | No clear effect |
| replace ERM with I-Mem decorrelation | DKT · Average of 4 real-world datasets | AUC: 0.6069 -> 0.6593 | Improved |
| replace ERM with I-Mem decorrelation | DKT · Synthetic FIIF alpha=0.6 | AUC: 0.6843 -> 0.7755 | Improved |
| replace ERM with I-Mem on strong-spurious FIIF | DKT · Synthetic FIIF alpha=0.9 | MAE: 0.0950 -> 0.0941 | No clear effect |
| replace ERM with I-Mem decorrelation | DKT · Synthetic FIIF alpha=0.6 | spurious/invariant importance: reduces dependency on spurious features by 16.9%, increases attention to invariant features by 275.1% | Improved |
| replace ERM with I-Mem on Anki | DKT · Anki | MAE: 0.4475 -> 0.3654 | Mixed |
Closest papers by meaning
- Diagnosing and Rectifying Fake OOD Invariance: A Restructured Causal Approach (AAAI 2024 · similarity 0.85)
- On the Memorization of Consistency Distillation for Diffusion Models (NeurIPS 2026 · similarity 0.84)
- Contrastive Consolidation of Top-Down Modulations Achieves Sparsely Supervised Continual Learning (NeurIPS 2025 · similarity 0.84)
- A mathematical theory of balancing relational generalization and memorization (NeurIPS 2026 · similarity 0.83)
- Measuring Dejavu Memorization Efficiently (NeurIPS 2024 · similarity 0.83)
- Deep Reinforcement Learning with Time-Scale Invariant Memory (AAAI 2025 · similarity 0.83)
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.
