arXiv subjects

Yanzhao Li

Publications and source records attributed to Yanzhao Li.

6 recordsLinked to original sources

arXiv2026

HiFA4: Training-Free 4-bit FlashAttention on Ascend HIF4 NPUs for LLM Inference

We present HiFA4, a post-training operator-level design that executes both QK^T and PV in FlashAttention as 4-bit HIF4 Cube GEMMs for LLM inference on Ascend NPUs, while maintaining the online softmax state in FP16. To our knowledge, HiFA4 is the first Ascend-HIF4-targeted design of this kind evaluated on standard NLP benchmarks. HiFA4 combines two mechanisms. Smooth-QK applies a calibration-static per-channel equivalent rescaling to Q and K after RoPE, transferring quantization difficulty from K to Q without per-tile online reduction at inference. P-Reordering accumulates the softmax normalizer from the same quantized attention weights P_hat used in the PV GEMM, rather than from a higher-precision reconstruction. We show that this inconsistent formulation introduces a coherent output-scaling error, and validate the effect on a Qwen3-8B Layer-0 MMLU trace, where all 3.6M measured attention tiles exhibit net probability-mass loss with median epsilon_bar = -0.064. P-Reordering also allows the normalizer to be fused into the PV Cube GEMM. Across five LLMs, HiFA4 reduces quantization-induced decision drift. On Qwen3-8B, it recovers 37.5% of the accuracy gap introduced by direct HIF4 quantization, narrows the sample-weighted accuracy loss from 1.12 pp to 0.70 pp, reduces BF16-inconsistent MMLU predictions from 16.3% to 8.2%, and cuts MMLU accuracy regressions by 57% (1071 to 465). On Gemma2-9B, mild smoothing keeps HiFA4 within 0.7 pp of BF16 while reducing MMLU regressions by 27%. On LLaMA3.1-8B, Mistral-7B, and Phi-4B, where Smooth-QK is disabled, P-Reordering with the adopted Q-Mean auxiliary still reduces full-set MMLU regressions by 41-52%. A preliminary instruction-scheduling analysis projects a 35.4% critical-path latency reduction relative to BF16 by fusing the softmax normalizer into the PV Cube GEMM; on-hardware validation is left to future work.

cs.LG↗

arXiv2026

Training-free sparse attention based on cumulative energy filtering

Sparse attention accelerates Diffusion Transformers (DiTs) for video generation by computing only the important tokens while skipping the rest. The token selection strategy is key to balancing sparsity and accuracy. We formulate the token filtering process as a dual-goal optimization problem: maximizing sparsity and minimizing accuracy degradation. Existing algorithms cannot fulfill both objectives simultaneously. For example, Top-p only considers the accuracy constraint, while Top-k maintains a fixed computational budget but loosens the accuracy constraint. This paper demonstrates that maintaining a fixed recall rate is sufficient for ensuring accuracy, whereas a fixed threshold is suboptimal for reducing computational cost. Therefore, we propose a dynamic thresholding scheme to improve sparsity while maintaining the same level of accuracy. Furthermore, our algorithm is deeply integrated with Flash Attention (FA), eliminating the need for any additional masking computation overhead. Experimental results on Wan 2.2 validate that, compared to the BLASST algorithm which is also integrated with FA, our dynamic thresholding strategy enhances sparsity from 61.42\% to 82\% with a VBench metric drop of less than 5\%. This results in an approximate 15\% in attention computation and a $1.61\times$ increase in computational efficiency, which is 1.18x higher than that of BLASST.

cs.CV↗

arXiv2026

OSC: Hardware Efficient W4A4 Quantization via Outlier Separation in Channel Dimension

While 4-bit quantization is essential for high-throughput deployment of Large Language Models, activation outliers often lead to significant accuracy degradation due to the restricted dynamic range of low-bit formats. In this paper, we systematically investigate the spatial distribution of outliers and demonstrate a token-persistent structural clustering effect, where high-magnitude outliers consistently occupy fixed channels across tokens. Building on this insight, we propose OSC, a hardware-efficient framework for outlier suppression. During inference, OSC executes a dual-path computation consisting of a low-precision 4-bit General Matrix Multiplication (GEMM) path and a high-precision 16-bit branch GEMM path. Specifically, OSC uses an offline group-wise strategy to identify the channels where outliers are located and then performs structured sub-tensor extraction to coalesce these scattered activation channels into a compact dense tensor online. This mechanism implements outlier protection through regularized and high-throughput GEMM operations, achieving a seamless fit with modern 4-bit micro-scaling hardware. Furthermore, for the inputs of W2 where outlier clustering is less pronounced, we integrate a fallback strategy to FP8. Evaluation on Qwen3-8B and Qwen3-30B restricts the average accuracy drop to 2.19 and 1.12 points, respectively. Notably, OSC is highly hardware-friendly, achieving a peak speedup of 1.78x over the W8A8 GEMM baseline on a modern AI accelerator.

cs.LG↗

arXiv2026

VFA: Relieving Vector Operations in Flash Attention with Global Maximum Pre-computation

FlashAttention-style online softmax enables exact attention computation with linear memory by streaming score tiles through on-chip memory and maintaining a running maximum and normalizer. However, as attention kernels approach peak tensor-core/cube-core throughput on modern accelerators, non-matmul components of online softmax -- especially per-tile rowmax and rowsum reductions and rescale chains -- can become vector or SIMD limited and dominate latency. This paper revisits FlashAttention and proposes Vector Relieved Flash Attention (VFA), a hardware-friendly method that reduces rowmax-driven updates of the running maximum while retaining the online-softmax structure. VFA initializes the running maximum via a cheap approximation from key-block representations, reorders key-block traversal to prioritize high-impact sink and local blocks, and freezes the maximum for remaining blocks to avoid repeated reductions and rescaling. We further integrate VFA with block-sparse skipping methods such as BLASST to form Vector Relieved Sparse Attention (VSA), which reduces both block count and per-block overhead. Notably, VFA and VSA completely avoid the conditional rescale operation in the update stage used in FA4.0. Extensive evaluations on benchmarks including MMLU and MATH500, together with attention statistics, verify our design: (i) sink and local reordering stabilizes the running maximum early; (ii) simple Q and K block summaries fail due to intra-block heterogeneity; (iii) m-initialization is required when maxima appear in middle blocks. Overall, VFA and VSA efficiently alleviate online-softmax reduction bottlenecks without performance loss. Compared to the C16V32 baseline, C8V32, C4V32 and C4V16 achieve nearly two times speedup on modern hardware while hitting the vector bottleneck. With upcoming architecture improvements, C4V16 will deliver six times speedup by enhancing exponent capacity.

cs.LG↗

arXiv2021

Optimal exit decision of venture capital under time-inconsistent preferences

This paper proposes two kinds of time-inconsistent preferences (i.e. time flow inconsistency and critical time point inconsistency) to further advance the research on the exit decision of venture capital. Time-inconsistent preference, different from time-consistent preference, assumes that decision makers prefer recent returns rather than future returns. Based on venture capitalists' understanding of future preferences, we consider four types of venture capitalists, namely time-consistent venture capitalists, venture capitalists who only realize critical time point inconsistency, naive venture capitalists and sophisticated venture capitalists, of which the latter three are time-inconsistent. All types of time-inconsistent venture capitalists are aware of critical time point inconsistency. Naive venture capitalists misunderstand time flow inconsistency while sophisticated ones understand it correctly. We propose an optimal exit timing of venture capital model. Then we derive and compare the above four types of venture capitalists' exit thresholds. The main results are as follows: (1) all types of time-inconsistent venture capitalists tend to exit earlier than time-consistent venture capitalists. (2) The longer the expire date are, the more likely venture capitalists are to delay the exit, but the delay degree decreases successively (venture capitalists who only realize critical time point inconsistency > naive venture capitalists > sophisticated venture capitalists).

q-fin.MF↗

arXiv2021

Tourism Demand Forecasting: An Ensemble Deep Learning Approach

The availability of tourism-related big data increases the potential to improve the accuracy of tourism demand forecasting, but presents significant challenges for forecasting, including curse of dimensionality and high model complexity. A novel bagging-based multivariate ensemble deep learning approach integrating stacked autoencoders and kernel-based extreme learning machines (B-SAKE) is proposed to address these challenges in this study. By using historical tourist arrival data, economic variable data and search intensity index (SII) data, we forecast tourist arrivals in Beijing from four countries. The consistent results of multiple schemes suggest that our proposed B-SAKE approach outperforms benchmark models in terms of level accuracy, directional accuracy and even statistical significance. Both bagging and stacked autoencoder can effectively alleviate the challenges brought by tourism big data and improve the forecasting performance of the models. The ensemble deep learning model we propose contributes to tourism forecasting literature and benefits relevant government officials and tourism practitioners.

stat.AP↗