Tag
该论文提出一种基于双向音频-文本交叉注意力的多模态框架，用于视频级别的矛盾与犹豫状态识别，在BAH数据集上取得77.95%的A/H F1分数。
HierDTA 是一个层次结构感知的药物-靶点亲和力预测框架，将药物表示为原子-基团异构图并结合蛋白质残基交互图与交叉注意力机制，在 Davis 和 KIBA 数据集上取得领先性能，并在冷启动场景下展现出良好泛化能力。
NVIDIA 的论文提出一种通过交叉注意力连接轻量 3D 运动模型与预训练视频生成器的方法，可为任意骨骼骨架同步生成视频与可回放的 3D 运动，并估计相机以实现运动与场景配准，支持视频与运动的双向转换。
Gen4P 将预训练视频生成器改造为前馈式 4D 感知器，通过单次编码与按需交叉注意力解码，从单目视频中预测几何与运动，并在 3D 点跟踪、视频深度和相机位姿估计上达到 SOTA 或有竞争力的性能。
BridgeWAM 通过潜在桥接查询（LBQs）解耦视频与动作专家，提升世界行动模型的动作预测效率与泛化能力，在 LIBERO 上达到 98.25% 成功率，同时将推理 FLOPs 降低 28.72%。
This paper introduces a Gated Cross Attention (GCA) framework that leverages LLM-generated explanations as stable background knowledge to detect fake news that has been rewritten under different emotional framings while preserving the underlying factual claims, showing improvements on PolitiFact, GossipCop, and LUN datasets.
A research project that fine-tunes the Aria piano transformer on twelve jazz pianists using gated cross-attention conditioning on learned pianist embeddings, achieving 70% intended-pianist attribution on generated continuations versus 37% without conditioning. The work, inspired by Dick Hyman's 'In the Styles of… The Great Jazz Pianists', is presented at ISMIR 2026 in Abu Dhabi.
GATE-ST proposes a gene-aware text-image encoder that integrates text-encoded gene summaries with histology image embeddings via cross-attention to improve spatial gene expression prediction, outperforming image-only and random-embedding baselines in pathology imaging benchmarks.
SMILESGNN introduces a multimodal architecture combining SMILES transformers and graph neural networks with cross-attention for interpretable drug toxicity prediction, achieving competitive performance on benchmarks like ClinTox and Tox21 with minimal parameters.
ROOSTER is a shared module that learns alignment between condition and target sequences for time-series forecasting and PPG-to-vital-sign reconstruction, achieving superior performance across multiple benchmarks.
The paper introduces RecCAR, a regularization method to address the reciprocal correspondence gap in joint multimodal diffusion transformers, improving performance in video generation tasks.
EnSol is an environment-aware graph neural network that predicts molecular solubility by representing solutes and solvents as graphs and using cross-attention to model interactions, with probabilistic outputs to capture temperature effects and experimental uncertainty. It achieves state-of-the-art performance on benchmark datasets, validated experimentally.
An interactive diagram comparing self-attention and cross-attention mechanisms in AI models, published as part of an educational library on attention.
QueryFormer is a unified transformer architecture that won the KDD Cup 2026 Tencent UniRec Challenge for post-click conversion rate prediction, focusing on query generation and efficient scaling.
This paper proposes Time–Frequency Geometric Cross-Attention (TFGCA), a drop-in module for chunked vision-language-action models that improves action trajectory prediction by decomposing chunks into time-frequency representations and capturing geometric relationships, resulting in significant performance gains on benchmarks and real-robot tasks.
RelightFormer introduces a feed-forward generative Transformer for direct single- and multi-view image relighting, using cross-attention for illumination injection and permutation-invariant encodings for unordered views, trained on a massive synthetic dataset to achieve state-of-the-art visual quality.
This paper investigates semantic leakage in audio-video diffusion models through the 'attention triangle' of cross-attention mechanisms, and presents methods to enhance semantic grounding during generation.
This paper introduces addressable and cardinality-preserving global memory for message-passing neural networks via cross-attention slots, addressing the finite-capacity bottleneck of virtual nodes and improving performance on multiplicity-aware tasks.
This paper proposes a dual-modal Vision Cross-Attention architecture for reaction yield prediction, fusing tabular physical-organic data with 2D molecular topologies, and demonstrates that a generic computer vision backbone can outperform purely quantum-based baselines.
Introduces SyRuP, a decoding-time framework that trains a cross-attention reward head to produce token-level adherence scores for system prompts, improving LLM following of complex prompts without model tuning.