SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs

Attention is the cornerstone of modern Large Language Models (LLMs). Yet its quadratic complexity limits the efficiency and scalability of LLMs, especially for those with a long-context window. A promising approach addressing this limitation is to leverage the sparsity in attention. However, existing sparsity-based solutions predominantly rely on predefined patterns or heuristics to approximate sparsity. This practice falls short to fully capture the dynamic nature of attention sparsity in language-based tasks. This paper argues that attention sparsity should be learned rather than predefined. To this end, we design SeerAttention, a new Attention mechanism that augments the conventional attention with a learnable gate that adaptively selects significant blocks in an attention map and deems the rest blocks sparse. Such block-level sparsity effectively balances accuracy and speedup. To enable efficient learning of the gating network, we develop a customized FlashAttention implementation that extracts the block-level ground truth of attention map with minimum overhead. SeerAttention not only applies to post-training, but also excels in long-context fine-tuning. Our results show that at post-training stages, SeerAttention significantly outperforms state-of-the-art static or heuristic-based sparse attention methods, while also being more versatile and flexible to adapt to varying context lengths and sparsity ratios. When applied to long-context fine-tuning with YaRN, SeerAttention can achieve a remarkable 90% sparsity ratio at a 32k context length with minimal perplexity loss, offering a 5.67x speedup over FlashAttention-2.

AdaRC: Mitigating Graph Structure Shifts during Test-Time

Powerful as they are, graph neural networks (GNNs) are known to be vulnerable to distribution shifts. Recently, test-time adaptation (TTA) has attracted attention due to its ability to adapt a pre-trained model to a target domain without re-accessing the source domain. However, existing TTA algorithms are primarily designed for attribute shifts in vision tasks, where samples are independent. These methods perform poorly on graph data that experience structure shifts, where node connectivity differs between source and target graphs. We attribute this performance gap to the distinct impact of node attribute shifts versus graph structure shifts: the latter significantly degrades the quality of node representations and blurs the boundaries between different node categories. To address structure shifts in graphs, we propose AdaRC, an innovative framework designed for effective and efficient adaptation to structure shifts by adjusting the hop-aggregation parameters in GNNs. To enhance the representation quality, we design a prediction-informed clustering loss to encourage the formation of distinct clusters for different node categories. Additionally, AdaRC seamlessly integrates with existing TTA algorithms, allowing it to handle attribute shifts effectively while improving overall performance under combined structure and attribute shifts. We validate the effectiveness of AdaRC on both synthetic and real-world datasets, demonstrating its robustness across various combinations of structure and attribute shifts.

LUT Tensor Core: Lookup Table Enables Efficient Low-Bit LLM Inference Acceleration

As large language model (LLM) inference demands ever-greater resources, there is a rapid growing trend of using low-bit weights to shrink memory usage and boost inference efficiency. However, these low-bit LLMs introduce the need for mixed-precision matrix multiplication (mpGEMM), which is a crucial yet under-explored operation that involves multiplying lower-precision weights with higher-precision activations. Unfortunately, current hardware does not natively support mpGEMM, resulting in indirect and inefficient dequantization-based implementations. To address the mpGEMM requirements in low-bit LLMs, we explored the lookup table (LUT)-based approach for mpGEMM. However, a conventional LUT implementation falls short of its potential. To fully harness the power of LUT-based mpGEMM, we introduce LUT Tensor Core, a software-hardware co-design optimized for low-bit LLM inference. Specifically, we introduce software-based operator fusion and table symmetrization techniques to optimize table precompute and table storage, respectively. Then, LUT Tensor Core proposes the hardware design featuring an elongated tiling shape design to enhance table reuse and a bit-serial design to support various precision combinations in mpGEMM. Moreover, we design an end-to-end compilation stack with new instructions for LUT-based mpGEMM, enabling efficient LLM compilation and optimizations. The evaluation on low-bit LLMs (e.g., BitNet, LLAMA) shows that LUT Tensor Core achieves more than a magnitude of improvements on both compute density and energy efficiency.

AIM: Attributing, Interpreting, Mitigating Data Unfairness

Data collected in the real world often encapsulates historical discrimination against disadvantaged groups and individuals. Existing fair machine learning (FairML) research has predominantly focused on mitigating discriminative bias in the model prediction, with far less effort dedicated towards exploring how to trace biases present in the data, despite its importance for the transparency and interpretability of FairML. To fill this gap, we investigate a novel research problem: discovering samples that reflect biases/prejudices from the training data. Grounding on the existing fairness notions, we lay out a sample bias criterion and propose practical algorithms for measuring and countering sample bias. The derived bias score provides intuitive sample-level attribution and explanation of historical bias in data. On this basis, we further design two FairML strategies via sample-bias-informed minimal data editing. They can mitigate both group and individual unfairness at the cost of minimal or zero predictive utility loss. Extensive experiments and analyses on multiple real-world datasets demonstrate the effectiveness of our methods in explaining and mitigating unfairness. Code is available at https://github.com/ZhiningLiu1998/AIM.

Discrete-state Continuous-time Diffusion for Graph Generation

Graph is a prevalent discrete data structure, whose generation has wide applications such as drug discovery and circuit design. Diffusion generative models, as an emerging research focus, have been applied to graph generation tasks. Overall, according to the space of states and time steps, diffusion generative models can be categorized into discrete-/continuous-state discrete-/continuous-time fashions. In this paper, we formulate the graph diffusion generation in a discrete-state continuous-time setting, which has never been studied in previous graph diffusion models. The rationale of such a formulation is to preserve the discrete nature of graph-structured data and meanwhile provide flexible sampling trade-offs between sample quality and efficiency. Analysis shows that our training objective is closely related to generation quality, and our proposed generation framework enjoys ideal invariant/equivariant properties concerning the permutation of node ordering. Our proposed model shows competitive empirical performance against state-of-the-art graph generation solutions on various benchmarks and, at the same time, can flexibly trade off the generation quality and efficiency in the sampling phase.

Allo: A Programming Model for Composable Accelerator Design

Special-purpose hardware accelerators are increasingly pivotal for sustaining performance improvements in emerging applications, especially as the benefits of technology scaling continue to diminish. However, designers currently lack effective tools and methodologies to construct complex, high-performance accelerator architectures in a productive manner. Existing high-level synthesis (HLS) tools often require intrusive source-level changes to attain satisfactory quality of results. Despite the introduction of several new accelerator design languages (ADLs) aiming to enhance or replace HLS, their advantages are more evident in relatively simple applications with a single kernel. Existing ADLs prove less effective for realistic hierarchical designs with multiple kernels, even if the design hierarchy is flattened. In this paper, we introduce Allo, a composable programming model for efficient spatial accelerator design. Allo decouples hardware customizations, including compute, memory, communication, and data type from algorithm specification, and encapsulates them as a set of customization primitives. Allo preserves the hierarchical structure of an input program by combining customizations from different functions in a bottom-up, type-safe manner. This approach facilitates holistic optimizations that span across function boundaries. We conduct comprehensive experiments on commonly-used HLS benchmarks and several realistic deep learning models. Our evaluation shows that Allo can outperform state-of-the-art HLS tools and ADLs on all test cases in the PolyBench. For the GPT2 model, the inference latency of the Allo generated accelerator is 1.7x faster than the NVIDIA A100 GPU with 5.4x higher energy efficiency, demonstrating the capability of Allo to handle large-scale designs.

Hierarchical Multi-Marginal Optimal Transport for Network Alignment

Finding node correspondence across networks, namely multi-network alignment, is an essential prerequisite for joint learning on multiple networks. Despite great success in aligning networks in pairs, the literature on multi-network alignment is sparse due to the exponentially growing solution space and lack of high-order discrepancy measures. To fill this gap, we propose a hierarchical multi-marginal optimal transport framework named HOT for multi-network alignment. To handle the large solution space, multiple networks are decomposed into smaller aligned clusters via the fused Gromov-Wasserstein (FGW) barycenter. To depict high-order relationships across multiple networks, the FGW distance is generalized to the multi-marginal setting, based on which networks can be aligned jointly. A fast proximal point method is further developed with guaranteed convergence to a local optimum. Extensive experiments and analysis show that our proposed HOT achieves significant improvements over the state-of-the-art in both effectiveness and scalability.

Ensuring User-side Fairness in Dynamic Recommender Systems

User-side group fairness is crucial for modern recommender systems, as it aims to alleviate performance disparity between groups of users defined by sensitive attributes such as gender, race, or age. We find that the disparity tends to persist or even increase over time. This calls for effective ways to address user-side fairness in a dynamic environment, which has been infrequently explored in the literature. However, fairness-constrained re-ranking, a typical method to ensure user-side fairness (i.e., reducing performance disparity), faces two fundamental challenges in the dynamic setting: (1) non-differentiability of the ranking-based fairness constraint, which hinders the end-to-end training paradigm, and (2) time-inefficiency, which impedes quick adaptation to changes in user preferences. In this paper, we propose FAir Dynamic rEcommender (FADE), an end-to-end framework with fine-tuning strategy to dynamically alleviate performance disparity. To tackle the above challenges, FADE uses a novel fairness loss designed to be differentiable and lightweight to fine-tune model parameters to ensure both user-side fairness and high-quality recommendations. Via extensive experiments on the real-world dataset, we empirically demonstrate that FADE effectively and efficiently reduces performance disparity, and furthermore, FADE improves overall recommendation quality over time compared to not using any new data.

Topological Augmentation for Class-Imbalanced Node Classification

Class imbalance is prevalent in real-world node classification tasks and often biases graph learning models toward majority classes. Most existing studies root from a node-centric perspective and aim to address the class imbalance in training data by node/class-wise reweighting or resampling. In this paper, we approach the source of the class-imbalance bias from an under-explored topology-centric perspective. Our investigation reveals that beyond the inherently skewed training class distribution, the graph topology also plays an important role in the formation of predictive bias: we identify two fundamental challenges, namely ambivalent and distant message-passing, that can exacerbate the bias by aggravating majority-class over-generalization and minority-class misclassification. In light of these findings, we devise a lightweight topological augmentation method ToBA to dynamically rectify the nodes influenced by ambivalent/distant message-passing during graph learning, so as to mitigate the class-imbalance bias. We highlight that ToBA is a model-agnostic, efficient, and versatile solution that can be seamlessly combined with and further boost other imbalance-handling techniques. Systematic experiments validate the superior performance of ToBA in both promoting imbalanced node classification and mitigating the prediction bias between different classes.