Quartet: Native FP4 Training Can Be Optimal for Large Language Models

The rapid advancement of large language models (LLMs) has been paralleled by unprecedented increases in computational demands, with training costs for state-of-the-art models doubling every few months. Training models directly in low-precision arithmetic offers a solution, by improving both computational throughput and energy efficiency. Specifically, NVIDIA's recent Blackwell architecture facilitates extremely low-precision operations, specifically FP4 variants, promising substantial efficiency gains. Yet, current algorithms for training LLMs in FP4 precision face significant accuracy degradation and often rely on mixed-precision fallbacks. In this paper, we systematically investigate hardware-supported FP4 training and introduce Quartet, a new approach enabling accurate, end-to-end FP4 training with all the major computations (in e.g. linear layers) being performed in low precision. Through extensive evaluations on Llama-type models, we reveal a new low-precision scaling law that quantifies performance trade-offs across varying bit-widths and allows us to identify a "near-optimal" low-precision training technique in terms of accuracy-vs-computation, called Quartet. We implement Quartet using optimized CUDA kernels tailored for NVIDIA Blackwell GPUs, and show that it can achieve state-of-the-art accuracy for FP4 precision, successfully training billion-scale models. Our method demonstrates that fully FP4-based training is a competitive alternative to standard-precision and FP8 training. Our code is available at https://github.com/IST-DASLab/Quartet.

UniDiffGrasp: A Unified Framework Integrating VLM Reasoning and VLM-Guided Part Diffusion for Open-Vocabulary Constrained Grasping with Dual Arms

Open-vocabulary, task-oriented grasping of specific functional parts, particularly with dual arms, remains a key challenge, as current Vision-Language Models (VLMs), while enhancing task understanding, often struggle with precise grasp generation within defined constraints and effective dual-arm coordination. We innovatively propose UniDiffGrasp, a unified framework integrating VLM reasoning with guided part diffusion to address these limitations. UniDiffGrasp leverages a VLM to interpret user input and identify semantic targets (object, part(s), mode), which are then grounded via open-vocabulary segmentation. Critically, the identified parts directly provide geometric constraints for a Constrained Grasp Diffusion Field (CGDF) using its Part-Guided Diffusion, enabling efficient, high-quality 6-DoF grasps without retraining. For dual-arm tasks, UniDiffGrasp defines distinct target regions, applies part-guided diffusion per arm, and selects stable cooperative grasps. Through extensive real-world deployment, UniDiffGrasp achieves grasp success rates of 0.876 in single-arm and 0.767 in dual-arm scenarios, significantly surpassing existing state-of-the-art methods, demonstrating its capability to enable precise and coordinated open-vocabulary grasping in complex real-world scenarios.

Deep Robust Reversible Watermarking

Robust Reversible Watermarking (RRW) enables perfect recovery of cover images and watermarks in lossless channels while ensuring robust watermark extraction in lossy channels. Existing RRW methods, mostly non-deep learning-based, face complex designs, high computational costs, and poor robustness, limiting their practical use. This paper proposes Deep Robust Reversible Watermarking (DRRW), a deep learning-based RRW scheme. DRRW uses an Integer Invertible Watermark Network (iIWN) to map integer data distributions invertibly, addressing conventional RRW limitations. Unlike traditional RRW, which needs distortion-specific designs, DRRW employs an encoder-noise layer-decoder framework for adaptive robustness via end-to-end training. In inference, cover image and watermark map to an overflowed stego image and latent variables, compressed by arithmetic coding into a bitstream embedded via reversible data hiding for lossless recovery. We introduce an overflow penalty loss to reduce pixel overflow, shortening the auxiliary bitstream while enhancing robustness and stego image quality. An adaptive weight adjustment strategy avoids manual watermark loss weighting, improving training stability and performance. Experiments show DRRW outperforms state-of-the-art RRW methods, boosting robustness and cutting embedding, extraction, and recovery complexities by 55.14\(\times\), 5.95\(\times\), and 3.57\(\times\), respectively. The auxiliary bitstream shrinks by 43.86\(\times\), with reversible embedding succeeding on 16,762 PASCAL VOC 2012 images, advancing practical RRW. DRRW exceeds irreversible robust watermarking in robustness and quality while maintaining reversibility.

Towards Robust Multi-UAV Collaboration: MARL with Noise-Resilient Communication and Attention Mechanisms

Efficient path planning for unmanned aerial vehicles (UAVs) is crucial in remote sensing and information collection. As task scales expand, the cooperative deployment of multiple UAVs significantly improves information collection efficiency. However, collaborative communication and decision-making for multiple UAVs remain major challenges in path planning, especially in noisy environments. To efficiently accomplish complex information collection tasks in 3D space and address robust communication issues, we propose a multi-agent reinforcement learning (MARL) framework for UAV path planning based on the Counterfactual Multi-Agent Policy Gradients (COMA) algorithm. The framework incorporates attention mechanism-based UAV communication protocol and training-deployment system, significantly improving communication robustness and individual decision-making capabilities in noisy conditions. Experiments conducted on both synthetic and real-world datasets demonstrate that our method outperforms existing algorithms in terms of path planning efficiency and robustness, especially in noisy environments, achieving a 78\% improvement in entropy reduction.

QuEST: Stable Training of LLMs with 1-Bit Weights and Activations

One approach to reducing the massive costs of large language models (LLMs) is the use of quantized or sparse representations for training or deployment. While post-training compression methods are very popular, the question of obtaining even more accurate compressed models by directly training over such representations, i.e., Quantization-Aware Training (QAT), is still open: for example, a recent study (arXiv:2411.04330v2) put the "optimal" bit-width at which models can be trained using QAT, while staying accuracy-competitive with standard FP16/BF16 precision, at 8-bits weights and activations. We advance this state-of-the-art via a new method called QuEST, which is Pareto-competitive with FP16, i.e., it provides better accuracy at lower model size, while training models with weights and activations in 4-bits or less. Moreover, QuEST allows stable training with 1-bit weights and activations. QuEST achieves this by improving two key aspects of QAT methods: (1) accurate and fast quantization of the (continuous) distributions of weights and activations via Hadamard normalization and MSE-optimal fitting; (2) a new trust gradient estimator based on the idea of explicitly minimizing the error between the noisy gradient computed over quantized states and the "true" (but unknown) full-precision gradient. Experiments on Llama-type architectures show that QuEST induces stable scaling laws across the entire range of hardware-supported precisions, and can be extended to sparse representations. We provide GPU kernel support showing that models produced by QuEST can be executed efficiently. Our code is available at https://github.com/IST-DASLab/QuEST.

PAID: A Framework of Product-Centric Advertising Image Design

In E-commerce platforms, a full advertising image is composed of a background image and marketing taglines. Automatic ad image design reduces human costs and plays a crucial role. For the convenience of users, a novel automatic framework named Product-Centric Advertising Image Design (PAID) is proposed in this work. PAID takes the product foreground image, required taglines, and target size as input and creates an ad image automatically. PAID consists of four sequential stages: prompt generation, layout generation, background image generation, and graphics rendering. Different expert models are trained to conduct these sub-tasks. A visual language model (VLM) based prompt generation model is leveraged to produce a product-matching background prompt. The layout generation model jointly predicts text and image layout according to the background prompt, product, and taglines to achieve the best harmony. An SDXL-based layout-controlled inpainting model is trained to generate an aesthetic background image. Previous ad image design methods take a background image as input and then predict the layout of taglines, which limits the spatial layout due to fixed image content. Innovatively, our PAID adjusts the stages to produce an unrestricted layout. To complete the PAID framework, we created two high-quality datasets, PITA and PIL. Extensive experimental results show that PAID creates more visually pleasing advertising images than previous methods.

SIESEF-FusionNet: Spatial Inter-correlation Enhancement and Spatially-Embedded Feature Fusion Network for LiDAR Point Cloud Semantic Segmentation

The ambiguity at the boundaries of different semantic classes in point cloud semantic segmentation often leads to incorrect decisions in intelligent perception systems, such as autonomous driving. Hence, accurate delineation of the boundaries is crucial for improving safety in autonomous driving. A novel spatial inter-correlation enhancement and spatially-embedded feature fusion network (SIESEF-FusionNet) is proposed in this paper, enhancing spatial inter-correlation by combining inverse distance weighting and angular compensation to extract more beneficial spatial information without causing redundancy. Meanwhile, a new spatial adaptive pooling module is also designed, embedding enhanced spatial information into semantic features for strengthening the context-awareness of semantic features. Experimental results demonstrate that 83.7% mIoU and 97.8% OA are achieved by SIESEF-FusionNet on the Toronto3D dataset, with performance superior to other baseline methods. A value of 61.1% mIoU is reached on the semanticKITTI dataset, where a marked improvement in segmentation performance is observed. In addition, the effectiveness and plug-and-play capability of the proposed modules are further verified through ablation studies.

Scalable Mechanistic Neural Networks

We propose Scalable Mechanistic Neural Network (S-MNN), an enhanced neural network framework designed for scientific machine learning applications involving long temporal sequences. By reformulating the original Mechanistic Neural Network (MNN) (Pervez et al., 2024), we reduce the computational time and space complexities from cubic and quadratic with respect to the sequence length, respectively, to linear. This significant improvement enables efficient modeling of long-term dynamics without sacrificing accuracy or interpretability. Extensive experiments demonstrate that S-MNN matches the original MNN in precision while substantially reducing computational resources. Consequently, S-MNN can drop-in replace the original MNN in applications, providing a practical and efficient tool for integrating mechanistic bottlenecks into neural network models of complex dynamical systems.

Enhancing Recommendation with Denoising Auxiliary Task

The historical interaction sequences of users plays a crucial role in training recommender systems that can accurately predict user preferences. However, due to the arbitrariness of user behavior, the presence of noise in these sequences poses a challenge to predicting their next actions in recommender systems. To address this issue, our motivation is based on the observation that training noisy sequences and clean sequences (sequences without noise) with equal weights can impact the performance of the model. We propose a novel self-supervised Auxiliary Task Joint Training (ATJT) method aimed at more accurately reweighting noisy sequences in recommender systems. Specifically, we strategically select subsets from users' original sequences and perform random replacements to generate artificially replaced noisy sequences. Subsequently, we perform joint training on these artificially replaced noisy sequences and the original sequences. Through effective reweighting, we incorporate the training results of the noise recognition model into the recommender model. We evaluate our method on three datasets using a consistent base model. Experimental results demonstrate the effectiveness of introducing self-supervised auxiliary task to enhance the base model's performance.

MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models

As inference on Large Language Models (LLMs) emerges as an important workload in machine learning applications, weight quantization has become a standard technique for efficient GPU deployment. Quantization not only reduces model size, but has also been shown to yield substantial speedups for single-user inference, due to reduced memory movement, with low accuracy impact. Yet, it remains open whether speedups are achievable also in \emph{batched} settings with multiple parallel clients, which are highly relevant for practical serving. It is unclear whether GPU kernels can be designed to remain practically memory-bound, while supporting the substantially increased compute requirements of batched workloads. This paper resolves this question positively by describing the design of Mixed-precision Auto-Regressive LINear kernels, called MARLIN. Concretely, given a model whose weights are compressed via quantization to, e.g., 4 bits per element, MARLIN shows that batchsizes up to 16-32 can be supported with close to maximum ($4\times$) quantization speedup, and larger batchsizes up to 64-128 with gradually decreasing, but still significant, acceleration. MARLIN accomplishes this via a combination of techniques, such as asynchronous memory access, complex task scheduling and pipelining, and bespoke quantization support. Our experiments show that MARLIN's near-optimal performance on individual LLM layers across different scenarios can also lead to end-to-end LLM inference speedups (of up to $2.8\times$) when integrated with the popular vLLM serving engine. Finally, MARLIN is extensible to further compression techniques, like NVIDIA 2:4 sparsity, leading to additional speedups.