Research map / Architectures and efficiency / Attention and transformers
Scout Before You Attend: Sketch-and-Walk Sparse Attention for Efficient LLM Inference
Hoang Anh Duy Le, Sahil Joshi, Zeyu Yang, Zhaozhuo Xu, Anshumali Shrivastava
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 |
|---|---|---|---|
| use Sketch&Walk prefill sparsity instead of dense attention | Llama-3.1-8B-Instruct · RULER prefill average | accuracy: 91.44 -> 91.35 | No clear effect |
| use Sketch&Walk instead of FlexPrefill | Llama-3.2-1B-Instruct · RULER prefill average | accuracy: 63.25 -> 67.59 | Improved |
| use Sketch&Walk instead of MInference | Llama-3.1-8B-Instruct · RULER prefill average | accuracy: 90.03 -> 91.35 | Improved |
| use Sketch&Walk instead of Quest for decoding | Llama-3.1-8B-Instruct · RULER decode average | accuracy: 89.06 -> 91.08 | Improved |
| use Sketch&Walk instead of Adamas for decoding | Llama-3.1-8B-Instruct · RULER decode average | accuracy: 90.74 -> 91.08 | Improved |
| apply end-to-end Sketch&Walk at 80% sparsity | Qwen2-7B-Instruct · LongBench AVG | accuracy: 40.85 -> 43.22 | Improved |
| apply end-to-end Sketch&Walk at 80% sparsity | Llama-3.1-8B-Instruct · LongBench AVG | accuracy: 47.95 -> 47.63 | No clear effect |
| apply end-to-end Sketch&Walk at 80% sparsity | Llama-3.2-1B-Instruct · LongBench AVG | accuracy: 29.75 -> 30.10 | No clear effect |
| reduce Hadamard sketch dimension to 16-32 (ablation) | Sketch&Walk · LongBench/RULER (3 datasets) | accuracy | No clear effect |
| increase walk degree (ablation) | Sketch&Walk · 3 datasets | accuracy | Improved |
| increase sparsity ratio (ablation) | Sketch&Walk · 3 datasets | accuracy | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FlexPrefill — FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference (ICLR 2025) · better on 4 of 4 shared results
- Quest — QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference (ICML 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention (NeurIPS 2024 · similarity 0.90)
- SparDA: Sparse Decoupled Attention for Efficient Long-Context LLM Inference (NeurIPS 2026 · similarity 0.89)
- Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration (ICML 2024 · similarity 0.88)
- SparQ Attention: Bandwidth-Efficient LLM Inference (ICML 2024 · similarity 0.88)
- Delta Attention: Fast and Accurate Sparse Attention Inference by Delta Correction (NeurIPS 2025 · similarity 0.88)
- QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference (ICML 2024 · similarity 0.88)
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.
