# TriAttention: Unified Sparse Mixed-Precision Attention for Long-Context LLM Inference

**Authors:** Hanzhi Zhang, Qiao Zhang, Qinglei Cao, Heng Fan, Yan Huang, Kewei Sha, Yunhe Feng  
**Published:** 5 Nov 2026  
**Links:** [arXiv](https://arxiv.org/abs/2611.07256) · [PDF](https://arxiv.org/pdf/2611.07256) · [Code](https://github.com/HanzhiZhang-Ulrica/TriAttention) · [This page](https://omen-bit.github.io/ai-paper-daily/paper/2611.07256.html)

## Key numbers

- **7.88×**: …TriAttention achieves end-to-end generation speedups of up to 7.88x over the Torch attention and 2.65x over…
- **2.65×**: …generation speedups of up to 7.88x over the Torch attention and 2.65x over the BF16 baseline.
- **4-bit**: tiles to skipping, 4-bit

## Abstract

Sparsity and quantization reduce the cost of attention by skipping interactions and lowering the precision of those retained. However, useful interactions differ in their precision requirements: some tolerate low-precision scores while remaining sensitive to removal. We therefore introduce TriAttention, which assigns query-key (QK) tiles to skipping, 4-bit, or 8-bit computation, using 4-bit to retain quantization-tolerant interactions at lower cost. To represent compactly, we encode each head's computation regions with two diagonal-distance boundaries, keeping policy storage independent of context length. Offline calibration then evaluates candidate boundary pairs using normalized projected-output errors and complete-tile execution costs. Using these measurements, a model-wide allocator selects each head's boundaries to minimize calibration error under a shared computation budget. During prefill, the selected boundaries generate precision-grouped traversals for a fused kernel that skips omitted tiles and combines contributions from both QK precisions into one normalized attention output. The quantized keys and values prepared during prefill are then reused in a compressed key-value (KV) cache for decoding, reducing repeated reads of prompt data. Experiments show that TriAttention achieves end-to-end generation speedups of up to 7.88x over the Torch attention and 2.65x over the BF16 baseline. Meanwhile, TriAttention maintains quality comparable to the original models on general-purpose and long-context tasks. These results show how a compact retention-and-precision policy translates local sensitivities into attention computation and repeated KV-cache access. The implementation is https://github.com/HanzhiZhang-Ulrica/TriAttention.

---

Notes written by AI from the paper, numbers copied from the abstract. Check the paper before you rely on them. From AI Papers Daily.