Research map / Architectures and efficiency / Attention and transformers
Training-Free Hashing-Based Attention via Binary Principal Components
Daohai Yu, Zhanpeng Zeng, Keyu Chen, Wenhao Li, Zhifeng Shen, Luxi Lin, Ruizhi Qiao, Xing Sun, Rongrong Ji
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 error-aware safeguard (EAS) at 2% budget (ablation) | BinaryPC (online projection) · InfiniteBench (Llama-3.1-8B) | average score: 41.30 -> 47.42 | Improved |
| add EAS to BinaryPC with offline calibration (ablation) | BinaryPC w/ OPC · InfiniteBench (Llama-3.1-8B) | average score: 46.93 -> 47.34 | Improved |
| increase EAS budget from 2% to 15% (ablation) | BinaryPC · InfiniteBench (Llama-3.1-8B) | average score: 47.42 to 47.63 | No clear effect |
| add offline projection calibration (OPC) | BinaryPC · Short-context LM-Eval (Llama-3-8B) | average accuracy: 59.31 -> 63.82 | Improved |
| increase hash length from 32 to 64 bits (ablation) | BinaryPC · NIAH (Llama-3.1-8B) | average score: 86.37 -> 90.18 | Improved |
| increase hash length from 64 to 128 bits (ablation) | BinaryPC · NIAH (Llama-3.1-8B) | no clear effect: 90.18 -> 90.00 | No clear effect |
| calibrate OPC on PG19 only instead of mixed corpora (ablation) | BinaryPC w/ OPC · LongBench | average score: 30.39 -> 30.43 | No clear effect |
| use data-aware binary principal components instead of random LSH projections | hashing-based sparse attention · RULER average (Llama-3.1-8B) | average score: 84.68 -> 85.57 | Improved |
| use 1K token budget instead of 2% (ablation) | BinaryPC · LongBench (Llama-3.1-8B) | average score: 49.50 -> 49.66 | Improved |
| use 2K budget instead of 2% on RULER (ablation) | BinaryPC · RULER average (Llama-3.1-8B) | average score: 85.57 -> 87.06 | Improved |
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.
- MagicPIG — MagicPIG: LSH Sampling for Efficient LLM Generation (ICLR 2025) · better on 6 of 6 shared results
- Quest (2K) — QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference (ICML 2024) · better on 3 of 3 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
- MagicPIG: LSH Sampling for Efficient LLM Generation (ICLR 2025 · similarity 0.91)
- Spotlight Attention: Towards Efficient LLM Generation via Non-linear Hashing-based KV Cache Retrieval (NeurIPS 2025 · similarity 0.91)
- Efficient Low Rank Attention for Long-Context Inference in Large Language Models (NeurIPS 2025 · similarity 0.88)
- More Value per Key: Asymmetric Sparse Attention for Faster LLM Decoding (NeurIPS 2026 · similarity 0.88)
- RocketKV: Accelerating Long-Context LLM Inference via Two-Stage KV Cache Compression (ICML 2025 · similarity 0.87)
- A Training-Free Sub-quadratic Cost Transformer Model Serving Framework with Hierarchically Pruned Attention (ICLR 2025 · similarity 0.87)
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.
