FoldAttention is presented, an additive formulation of softmax attention that fixes a finite reference $Z_i$ before scanning the KV cache and removes the determinism tax.
Abstract
Autoregressive decode repeatedly streams a growing KV cache, making attention a major cost at long context. Existing high-performance kernels use online softmax, which discovers a row's normalization reference as it scans keys. Earlier contributions therefore remain provisional and may require rescaling. We argue that the reference need not be discovered: softmax is invariant to a common shift, so the reference only has to keep the weights in range. We present FoldAttention, an additive formulation of softmax attention that fixes a finite reference $Z_i$ before scanning the KV cache. Each weight $2^{s_{ij}-Z_i}$ is then final when computed, so contributions add across disjoint key ranges and their quotient equals softmax attention in real arithmetic. We use this property to develop two techniques for Hopper decode: (1) final weights gate key and value reads before the bytes are fetched, and a per-call depth $T$ cuts keys below $2^{-T}$ while keeping their mass, and (2) additive partials compose split KV and shared-prefix cascades without rescaling. On H100 at $T=16$, FoldAttention decodes seven real-model generations 1.36-2.30$\times$ faster than the fastest BF16 baseline, and up to 3.09$\times$ faster across MHA and GQA shapes, at an error within 1.5% of the lowest BF16 error on six of the seven; reading every key, it is 1.14-1.30$\times$ faster at matched error. We validate on Qwen3-8B that a whole decode step is up to 1.46$\times$ faster while likelihood and long-context accuracy match those under BF16 kernels. The same principle makes the backward deterministic: CTAs round bounded partial gradients onto an integer grid declared before the reduction and add them in any order. FoldAttention thereby removes the determinism tax: its deterministic backward is up to 1.84$\times$ faster than deterministic FlashAttention-3/4 and 1.05$\times$ faster than the fastest nondeterministic kernel.
On NVIDIA Blackwell B200, tensor-core throughput outpaces special-function exponential throughput by more than two orders of magnitude, exposing exponential evaluation in fused attention kernels. A pretrained Transformer, however, may not need it evaluated accurately at every element. We characterize what a pretrained...
Shan-Zhen Zhu, Mu-Yan Hu, Tomasz Kozlowski· 0 citations
On Blackwell, normalization and operand movement can limit attention kernels whose matrix multiplications are already deeply pipelined. We implement an FA4-style pipeline in Triton TLX with fixed-shift dense softmax and a saved inverse denominator for backward. The fixed shift removes recurrent accumulator corrections;...
Олександр Сташук, Hong-Tao Yu, J. Shah· 1 citation
AMEND predicts each block's BLASST verdict from margins audited at earlier steps, so the GPU fetches only predicted survivors while near-bank PIM units in high-bandwidth memory (HBM) concurrently score the omitted complement.
Zu-Xiong Tan, Will Wei-Jen Wang, Wei Shao et al.· 0 citations
The detachment mechanism of Forward-Forward (FF) is reinterpreted as a scheduling primitive: given a local objective, detaching a block's output removes downstream gradient dependencies, making its backward pass ready when its forward pass finishes.
This work shows that Sliding Window Attention (SWA) with sinks performs as well or better than post-trained Linear Attention models, and recommends switching to SWA instead of post-training linear models.
Alexia Jolicoeur-Martineau, R. Sukthanker, Pashmina Cameron et al.· 0 citations
This framework separates temporal alignment, plasticity, forgetting, and bounded rehearsal in recurrent sequence models, together with numerically stable positive-decay renormalization, to remain competitive in language modeling and improve length extrapolation on variable-digit addition.
Yi-Fan Zhang, Steve Ta, Jasper Zhang et al.· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026