Research map / Architectures and efficiency / Attention and transformers
Compact and Structurally Transparent Cervical Cytology with Geometry-Driven Features and Closed-Form Attention
Dichao Liu
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 local N/C ratio concept (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 95.3 -> 98.5 | Improved |
| add nonnegative constraint on concept fusion (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 95.9 -> 98.5 | Improved |
| add nuclear roughness concept (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 96.4 -> 98.5 | Improved |
| add orientation disorder concept (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 96.9 -> 98.5 | Improved |
| add chromatin heterogeneity concept (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 97.8 -> 98.5 | Improved |
| use 4 backbone blocks instead of 1 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 82.4 -> 98.5 | Improved |
| use 5 backbone blocks instead of 4 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 98.5 -> 97.1 | No clear effect |
| use attention lattice at 1/32 resolution instead of 1/4 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 25.8 -> 98.5 | Improved |
| use attention lattice at 1/32 resolution instead of 1/64 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 97.2 -> 98.5 | Improved |
| use residual step 0.5 instead of 0.2 or 0.8 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 97.8 / 97.0 -> 98.5 | Improved |
| use 9x9 local window instead of 11x11 (ablation) | GeoCEAN · SIPaKMeD | ACC (%): 96.8 -> 98.5 | Improved |
| use GeoCEAN instead of Swin-GMRS (strongest baseline) | cervical cytology classifier · DSCC | ACC (%): 89.5 -> 93.2 | Improved |
Closest papers by meaning
- SheafStain: Sheaf-Theoretic Schr\"odinger Bridge for Spatially and Biologically Coherent Virtual Staining (NeurIPS 2026 · similarity 0.81)
- CPathAgent: An Agent-based Foundation Model for Interpretable High-Resolution Pathology Image Analysis Mimicking Pathologists' Diagnostic Logic (NeurIPS 2025 · similarity 0.79)
- TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis (ICLR 2026 · similarity 0.79)
- Symb-xMIL: Symbolic Explanations for Multiple Instance Learning in Digital Pathology (NeurIPS 2026 · similarity 0.79)
- HEST-1k: A Dataset For Spatial Transcriptomics and Histology Image Analysis (NeurIPS 2024 · similarity 0.79)
- MorphoHELM: A Comprehensive Benchmark for Evaluating Representations for Microscopy-Based Morphology Assays (NeurIPS 2026 · similarity 0.79)
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.
