首页 > AI前沿 > Elastic Threshold Attention: Learned Contextual Sparsity for Long-Context Decoding

Elastic Threshold Attention: Learned Contextual Sparsity for Long-Context Decoding

arXiv机器学习 2026-09-17 07:23 7 阅读 查看原文

Massive KV caches can cause severe memory-bandwidth bottlenecks during long-context decoding.

Sparse attention methods mitigate this via selective loading, but that comes at a cost: rigid heuristics drop necessary context, leading to quality degradation.

We introduce Elastic Threshold Attention (ETA), an end-to-end trainable architecture that achieves hardware-accelerated decoding speed without sacrificing dense model quality.

ETA predicts dynamic, contextual thresholds directly from query representations, allowing the model to allocate dense-like context to difficult retrieval or reasoning steps while pruning routine tokens.

To learn this policy from scratch without representation collapse, ETA multiplicatively suppresses sub-threshold logits toward zero during training rather than deleting them.

Training against this smooth uniform attention floor provides a distributed probability reservoir that causes localized attention sinks on initial tokens to disappear.

It also enables the model to hard-prune uninformative KV blocks at inference time and absorb incidental tokens co-admitted by coarse GPU block selection.

As a result, a 1.45B pretrained ETA model rivals dense attention across language modeling, commonsense reasoning, and long-context needle retrieval at $\approx 85\%$ training sparsity and $\approx 38\%$ active decode density.

At inference time, we implement a custom decode kernel in Triton that screens KV blocks in $O(1)$ time using cached geometric-probabilistic bounds, delivering up to $2.5\times$ wall-clock decode speedups over FlashAttention-2 on sequences up to 512K tokens.

Finally, we introduce an offline calibration algorithm for domain-specific deployments that freezes per-head constant thresholds to eliminate predictor overhead, cutting attention compute by an additional $27\%$.