Research map / Architectures and efficiency / RNNs and normalization
SparseMAP: Differentiable Sparse Structured Inference
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use SparseMAP loss instead of Structured SVM loss | bi-LSTM dependency parser · UD dependency parsing zh | unlabeled attachment accuracy: 81.94 -> 84.03 | Helped |
| use SparseMAP loss instead of CRF loss | bi-LSTM dependency parser · UD dependency parsing vi | unlabeled attachment accuracy: 69.10 -> 69.71 | Helped |
| use SparseMAP loss instead of Structured SVM loss | bi-LSTM dependency parser · UD dependency parsing ja | unlabeled attachment accuracy: 96.24 -> 96.04 | Hurt |
| add margin cost to SparseMAP loss | bi-LSTM dependency parser · UD dependency parsing en | unlabeled attachment accuracy: 86.90 -> 87.34 | Helped |
| add margin cost to SparseMAP loss | bi-LSTM dependency parser · UD dependency parsing zh | unlabeled attachment accuracy: 84.03 -> 82.63 | Hurt |
| add margin cost to SparseMAP loss | bi-LSTM dependency parser · UD dependency parsing vi | unlabeled attachment accuracy: 69.71 -> 70.87 | Helped |
| add margin cost to SparseMAP loss | bi-LSTM dependency parser · UD dependency parsing ro | unlabeled attachment accuracy: 87.35 -> 87.63 | Helped |
| add margin cost to SparseMAP loss | bi-LSTM dependency parser · UD dependency parsing ja | unlabeled attachment accuracy: 96.04 -> 96.03 | Hurt |
| replace softmax with sequential SparseMAP alignment | ESIM · MultiNLI | test accuracy: 76.05 -> 75.54 | Hurt |
| replace softmax with matching SparseMAP alignment | ESIM · MultiNLI | test accuracy: 76.05 -> 76.13 | Helped |
| replace softmax with sequential SparseMAP alignment | ESIM · SNLI | test accuracy: 86.52 -> 86.62 | Helped |
| replace softmax with matching SparseMAP alignment | ESIM · SNLI | test accuracy: 86.52 -> 86.05 | Hurt |
Closest papers by meaning
- Dynamic Sparse Training with Structured Sparsity (ICLR 2024 · similarity 0.88)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.87)
- R-Sparse: Rank-Aware Activation Sparsity for Efficient LLM Inference (ICLR 2025 · similarity 0.87)
- The Sparsity Whisperer (NeurIPS 2026 · similarity 0.86)
- Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs (ICLR 2024 · similarity 0.86)
- Týr-the-Pruner: Structural Pruning LLMs via Global Sparsity Distribution Optimization (NeurIPS 2025 · similarity 0.86)
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.