Multi-Head LatentMoE and Head Parallel: Communication-Efficient and Deterministic MoE Parallelism

Large language models have transformed many applications but remain expensive to train. Sparse Mixture of Experts (MoE) addresses this through conditional computation, with Expert Parallel (EP) as the standard distributed training method. However, EP has three limitations: communication cost grows linearly with the number of activated experts $k$, load imbalance affects latency and memory usage, and data-dependent communication requires metadata exchange. We propose Multi-Head LatentMoE and Head Parallel (HP), a new architecture and parallelism achieving $O(1)$ communication cost regardless of $k$, completely balanced traffic, and deterministic communication, all while remaining compatible with EP. To accelerate Multi-Head LatentMoE, we propose IO-aware routing and expert computation. Compared to MoE with EP, Multi-Head LatentMoE with HP trains up to $1.61\times$ faster while having identical performance. With doubled granularity, it achieves higher overall performance while still being $1.11\times$ faster. Our method makes multi-billion-parameter foundation model research more accessible.

An All-MLP Sequence Modeling Architecture That Excels at Copying

Recent work demonstrated Transformers' ability to efficiently copy strings of exponential sizes, distinguishing them from other architectures. We present the Causal Relation Network (CausalRN), an all-MLP sequence modeling architecture that can match Transformers on the copying task. Extending Relation Networks (RNs), we implemented key innovations to support autoregressive sequence modeling while maintaining computational feasibility. We discovered that exponentially-activated RNs are reducible to linear time complexity, and pre-activation normalization induces an infinitely growing memory pool, similar to a KV cache. In ablation study, we found both exponential activation and pre-activation normalization are indispensable for Transformer-level copying. Our findings provide new insights into what actually constitutes strong in-context retrieval.

Generalizing and Improving Jacobian and Hessian Regularization

Jacobian and Hessian regularization aim to reduce the magnitude of the first and second-order partial derivatives with respect to neural network inputs, and they are predominantly used to ensure the adversarial robustness of image classifiers. In this work, we generalize previous efforts by extending the target matrix from zero to any matrix that admits efficient matrix-vector products. The proposed paradigm allows us to construct novel regularization terms that enforce symmetry or diagonality on square Jacobian and Hessian matrices. On the other hand, the major challenge for Jacobian and Hessian regularization has been high computational complexity. We introduce Lanczos-based spectral norm minimization to tackle this difficulty. This technique uses a parallelized implementation of the Lanczos algorithm and is capable of effective and stable regularization of large Jacobian and Hessian matrices. Theoretical justifications and empirical evidence are provided for the proposed paradigm and technique. We carry out exploratory experiments to validate the effectiveness of our novel regularization terms. We also conduct comparative experiments to evaluate Lanczos-based spectral norm minimization against prior methods. Results show that the proposed methodologies are advantageous for a wide range of tasks.

Conceptual Text Region Network: Cognition-Inspired Accurate Scene Text Detection

Segmentation-based methods are widely used for scene text detection due to their superiority in describing arbitrary-shaped text instances. However, two major problems still exist: 1) current label generation techniques are mostly empirical and lack theoretical support, discouraging elaborate label design; 2) as a result, most methods rely heavily on text kernel segmentation which is unstable and requires deliberate tuning. To address these challenges, we propose a human cognition-inspired framework, termed, Conceptual Text Region Network (CTRNet). The framework utilizes Conceptual Text Regions (CTRs), which is a class of cognition-based tools inheriting good mathematical properties, allowing for sophisticated label design. Another component of CTRNet is an inference pipeline that, with the help of CTRs, completely omits the need for text kernel segmentation. Compared with previous segmentation-based methods, our approach is not only more interpretable but also more accurate. Experimental results show that CTRNet achieves state-of-the-art performance on benchmark CTW1500, Total-Text, MSRA-TD500, and ICDAR 2015 datasets, yielding performance gains of up to 2.0%. Notably, to the best of our knowledge, CTRNet is among the first detection models to achieve F-measures higher than 85.0% on all four of the benchmarks, with remarkable consistency and stability.