The paper introduces Multi-Stream Attention Residuals (MSAR), an extension of Attention Residuals that applies multi-head depth attention inspired by multi-head sequence attention. MSAR uses multi-scale causal convolutions to provide diverse contextual views for each head, then adaptively mixes the retrieved streams into a single-stream input, avoiding persistent multi-stream storage. Across 12 downstream benchmarks at 18B and 28B MoE scales, MSAR outperforms prior residual variants and achieves a 1.58× compute speedup over the standard residual baseline.
By Xiaohan Qin, Xiangdong Zhang, Tuo Dai, Sunan Zou, Huaijin Wu, Yebin Yang, Zhuo Xia, Lin Yao, Yuliang Liu, Junchi Yan
Multi-Stream Attention Residuals (MSAR) extends the Attention Residuals (AttnRes) mechanism by introducing multi-head depth attention, inspired by multi-head attention along the sequence axis. MSAR generates diverse streams from sequence context via multi-scale causal convolutions, then adaptively mixes them to produce a single-stream layer input, enabling multi-stream computation without extra storage or communication. Across 12 downstream benchmarks at 18B and 28B MoE scales, MSAR consistently outperforms prior residual variants and achieves a 1.58× iso-loss compute speedup over the standard residual baseline.
The paper introduces CATTLE, a transfer learning framework for disjoint tabular datasets that eliminates the need for shared features by leveraging generalized context learned through transformer projection weights. By using key, value, and query weights from source and target domains, CATTLE performs cross‑domain attention transfer in a data‑agnostic manner. Experiments on ten source‑target pairs demonstrate that CATTLE outperforms nine state‑of‑the‑art baselines, achieving the best average rank (2.9) and a 3.7% AUROC improvement.
By Kazi F. Akhter, Ibna Kowsar, Manar D. Samad
arXiv:2607. 27303v1 Announce Type: new Abstract: Temporal heterogeneous graphs offer a natural abstraction for dynamic relational systems in which diverse node and relation types co-exist and evolve over time.
By Yixin Peng, Diego Collarana, Er Jin, Stefan Decker
The paper investigates how attention dynamics evolve across recurrent depth in language models, finding that attention support stabilizes early while hidden states and outputs take longer. It proposes WISE, a training‑free method that uses full attention in early steps and then reuses the discovered sparse working set for later steps, preserving performance on multi‑hop QA tasks. Experiments show that WISE maintains quality up to 2K context, offers measurable speedups, and highlights the importance of recurrent discovery of attention support.
By Ke Wan, Chen Chen
arXiv:2512. 12477v2 Announce Type: replace Abstract: Estimating node importance in heterogeneous knowledge graphs is a fundamental problem underlying recommendation, search, and knowledge decision systems.
By Jiawen Chen, Yanyan He, Qi Shao, Mengli Wei, Duxin Chen, Wenwu Yu, Yanlong Zhao
arXiv:2604. 17324v2 Announce Type: replace-cross Abstract: Global self-attention drives modern graph transformers, yet the softmax at its core imposes a structural constraint rarely examined directly: every attention row is non-negative and sums to one, so each per-head output is a mass-conserving convex combination of value vectors.
By Yang Liu, Dongxin Guo, Tom Zheng, Siu Ming Yiu, Liam Ning, Jikun Wu
The paper investigates whether deep transformer layers require context from the residual stream to compute value vectors. It finds that allowing deeper layers to use a context‑free value vector—preserving original token information—significantly improves performance, and adding context afterward yields little extra benefit. The authors introduce Bank of Values (BoV), a lookup table of token‑specific value vectors for the last third of layers, which reduces compute and memory while matching or surpassing prior methods on large models.
By Muyu He, Yuchen Liu, Qingya Huang, Li Zhang
The paper introduces Higher-Order Modular Attention (HOMA), a new attention mechanism that combines standard pairwise self‑attention with an explicit triadic attention pathway. HOMA uses overlapping blocks, local windows, and a low‑rank projection to make triadic interactions tractable. Experiments on controlled PARITY and MATCH3 tasks, as well as TAPE benchmarks, show that HOMA matches or outperforms matched pairwise and purely triadic baselines, especially when dependencies extend beyond triadic order, and it often converges faster and uses parameters more efficiently.
By Shirin Amiraslani, Xin Gao
arXiv:2607. 05017v1 Announce Type: cross Abstract: The performance of deep learning models crucially depends on the settings of hyperparameters like learning rate, initialization scale, and weight decay.
By Gage DeZoort, Boris Hanin
The paper introduces OASIS, a method designed to stabilize dual‑normalized attention‑residual architectures by employing explicit null routing and token‑to‑depth null coupling. OASIS mitigates attention sinks and activation outliers, improving low‑bit quantization performance across several language‑model backbones. Empirical results show significant reductions in attention norms and perplexity, with notable gains on long‑context benchmarks.
By Haozheng Luo, Haoran Dai, Jingyuan Huang, Shaoyang Zhang, Xi Chen, Eric Hanchen Jiang, Yijiang Li, Chenghao Qiu, Chenwei Xu, Zhenyu Pan, Haotian Zhang, Binghui Wang, Yan Chen
The paper introduces STD, a hierarchical token pruning framework for Large Vision‑Language Models that aligns pruning strategies with the functional roles of different network stages. By using high‑frequency spectral analysis in shallow layers, Gaussian‑smoothed attention in intermediate layers, and a stability‑adaptive trigger in deep layers, STD preserves essential visual information while aggressively reducing token counts. Experiments demonstrate that STD outperforms existing pruning methods, achieving up to 94.4% token reduction and a 3.9× speed‑up on LLaVA‑NeXT‑7B.
By Shuo Zhang, Jintao Tong, Yixiong Zou, Yuhua Li, Ruixuan Li