arXiv subjects

Jue Wang

Publications and source records attributed to Jue Wang.

At least 55 records · Page 3Linked to original sources

arXiv2025

Beat the long tail: Distribution-Aware Speculative Decoding for RL Training

Reinforcement learning(RL) post-training has become essential for aligning large language models (LLMs), yet its efficiency is increasingly constrained by the rollout phase, where long trajectories are generated token by token. We identify a major bottleneck:the long-tail distribution of rollout lengths, where a small fraction of long generations dominates wall clock time and a complementary opportunity; the availability of historical rollouts that reveal stable prompt level patterns across training epochs. Motivated by these observations, we propose DAS, a Distribution Aware Speculative decoding framework that accelerates RL rollouts without altering model outputs. DAS integrates two key ideas: an adaptive, nonparametric drafter built from recent rollouts using an incrementally maintained suffix tree, and a length aware speculation policy that allocates more aggressive draft budgets to long trajectories that dominate makespan. This design exploits rollout history to sustain acceptance while balancing base and token level costs during decoding. Experiments on math and code reasoning tasks show that DAS reduces rollout time up to 50% while preserving identical training curves, demonstrating that distribution-aware speculative decoding can significantly accelerate RL post training without compromising learning quality.

cs.LG↗

arXiv2025

Enlightenment Period Improving DNN Performance

The start of deep neural network training is characterized by a brief yet critical phase that lasts from the beginning of the training until the accuracy reaches approximately 50\%. During this phase, disordered representations rapidly transition toward ordered structure, and we term this phase the Enlightenment Period. Through theoretical modeling based on phase transition theory and experimental validation, we reveal that applying Mixup data augmentation during this phase has a dual effect: it introduces a Gradient Interference Effect that hinders performance, while also providing a beneficial Activation Revival Effect to restore gradient updates for saturated neurons. We further demonstrate that this negative interference diminishes as the sample set size or the model parameter size increases, thereby shifting the balance between these two effects. Based on these findings, we propose three strategies that improve performance by solely adjusting the training data distribution within this brief period: the Mixup Pause Strategy for small-scale scenarios, the Alpha Boost Strategy for large-scale scenarios with underfitting, and the High-Loss Removal Strategy for tasks where Mixup is inapplicable (e.g., time series and large language models). Extensive experiments show that these strategies achieve superior performance across diverse architectures such as ViT and ResNet on datasets including CIFAR and ImageNet-1K. Ultimately, this work offers a novel perspective on enhancing model performance by strategically capitalizing on the dynamics of the brief and crucial early stages of training. Code is available at https://anonymous.4open.science/r/code-A5F1/.

cs.LG↗

arXiv2025

MOOSE-Chem3: Toward Experiment-Guided Hypothesis Ranking via Simulated Experimental Feedback

Hypothesis ranking is vital for automated scientific discovery, especially in cost-intensive, throughput-limited natural science domains. Current methods focus on pre-experiment ranking, relying solely on language model reasoning without empirical feedback. We introduce experiment-guided ranking, which prioritizes hypotheses based on feedback from prior tests. Due to the impracticality of real experiments, we propose a simulator grounded in domain-specific concepts that models hypothesis performance as a function of similarity to a hidden ground truth, perturbed by noise. Validated against 124 hypotheses with experimentally reported outcomes, the simulator approximates real results with consistent trend alignment. Although deviations exist, they mimic wet-lab noise, promoting more robust ranking strategies. We frame experiment-guided ranking as a sequential decision-making problem and propose an in-context reinforcement learning (ICRL) framework. Our LLM-based policy decomposes hypotheses into functional elements, clusters them by mechanistic roles, and prioritizes recombinations based on feedback. Experiments show our approach significantly outperforms pre-experiment baselines and strong ablations. Our toolkit, comprising the simulator and ICRL framework, enables systematic research on experiment-guided ranking, with the policy serving as a strong proof of concept.

cs.CL↗

arXiv2025

Ultra High-Resolution Image Inpainting with Patch-Based Content Consistency Adapter

In this work, we present Patch-Adapter, an effective framework for high-resolution text-guided image inpainting. Unlike existing methods limited to lower resolutions, our approach achieves 4K+ resolution while maintaining precise content consistency and prompt alignment, two critical challenges in image inpainting that intensify with increasing resolution and texture complexity. Patch-Adapter leverages a two-stage adapter architecture to scale the diffusion model's resolution from 1K to 4K+ without requiring structural overhauls: (1) Dual Context Adapter learns coherence between masked and unmasked regions at reduced resolutions to establish global structural consistency; and (2) Reference Patch Adapter implements a patch-level attention mechanism for full-resolution inpainting, preserving local detail fidelity through adaptive feature fusion. This dual-stage architecture uniquely addresses the scalability gap in high-resolution inpainting by decoupling global semantics from localized refinement. Experiments demonstrate that Patch-Adapter not only resolves artifacts common in large-scale inpainting but also achieves state-of-the-art performance on the OpenImages and Photo-Concept-Bucket datasets, outperforming existing methods in both perceptual quality and text-prompt adherence.

cs.CV↗

arXiv2025

Maritime Communication in Evaporation Duct Environment with Ship Trajectory Optimization

In maritime wireless networks, the evaporation duct effect has been known as a preferable condition for long-range transmissions. However, how to effectively utilize the duct effect for efficient communication design is still open for investigation. In this paper, we consider a typical scenario of ship-to-shore data transmission, where a ship collects data from multiple oceanographic buoys, sails from one to another, and transmits the collected data back to a terrestrial base station during its voyage. A novel framework, which exploits priori information of the channel gain map in the presence of evaporation duct, is proposed to minimize the data transmission time and the sailing time by optimizing the ship's trajectory. To this end, a multi-objective optimization problem is formulated and is further solved by a dynamic population PSO-integrated NSGA-II algorithm. Through simulations, it is demonstrated that, compared to the benchmark scheme which ignores useful information of the evaporation duct, the proposed scheme can effectively reduce both the data transmission time and the sailing time.

eess.SP↗

arXiv2025

Staircase Streaming for Low-Latency Multi-Agent Inference

Recent advances in large language models (LLMs) opened up new directions for leveraging the collective expertise of multiple LLMs. These methods, such as Mixture-of-Agents, typically employ additional inference steps to generate intermediate outputs, which are then used to produce the final response. While multi-agent inference can enhance response quality, it can significantly increase the time to first token (TTFT), posing a challenge for latency-sensitive applications and hurting user experience. To address this issue, we propose staircase streaming for low-latency multi-agent inference. Instead of waiting for the complete intermediate outputs from previous steps, we begin generating the final response as soon as we receive partial outputs from these steps. Experimental results demonstrate that staircase streaming reduces TTFT by up to 93% while maintaining response quality.

cs.AI↗

arXiv2025

Investigation of Superdirectivity in Planar Holographic Arrays

This paper studies the superdirectivity characteristics of uniform rectangular arrays (URAs) for holographic multiple-input multiple-output systems. By establishing a mathematical directivity model for the URA, an analytical expression for the maximum directivity is derived. Accordingly, systematic analysis is performed in conjunction with numerical simulations. Results show that the directivity can be significantly enhanced via rational utilization of coupling effects. However, this enhancement yields diminishing returns when antenna spacings transition to deep sub-wavelength scales. This study provides a theoretical basis for the design of superdirective URAs and offers valuable insights for holographic array optimization in 5G/6G communication systems.

eess.SP↗

arXiv2025

Lossless Compression: A New Benchmark for Time Series Model Evaluation

The evaluation of time series models has traditionally focused on four canonical tasks: forecasting, imputation, anomaly detection, and classification. While these tasks have driven significant progress, they primarily assess task-specific performance and do not rigorously measure whether a model captures the full generative distribution of the data. We introduce lossless compression as a new paradigm for evaluating time series models, grounded in Shannon's source coding theorem. This perspective establishes a direct equivalence between optimal compression length and the negative log-likelihood, providing a strict and unified information-theoretic criterion for modeling capacity. Then We define a standardized evaluation protocol and metrics. We further propose and open-source a comprehensive evaluation framework TSCom-Bench, which enables the rapid adaptation of time series models as backbones for lossless compression. Experiments across diverse datasets on state-of-the-art models, including TimeXer, iTransformer, and PatchTST, demonstrate that compression reveals distributional weaknesses overlooked by classic benchmarks. These findings position lossless compression as a principled task that complements and extends existing evaluation for time series modeling.

cs.LG↗

arXiv2025

DPDEdit: Detail-Preserved Diffusion Models for Multimodal Fashion Image Editing

Fashion image editing is a crucial tool for designers to convey their creative ideas by visualizing design concepts interactively. Current fashion image editing techniques, though advanced with multimodal prompts and powerful diffusion models, often struggle to accurately identify editing regions and preserve the desired garment texture detail. To address these challenges, we introduce a new multimodal fashion image editing architecture based on latent diffusion models, called Detail-Preserved Diffusion Models (DPDEdit). DPDEdit guides the fashion image generation of diffusion models by integrating text prompts, region masks, human pose images, and garment texture images. To precisely locate the editing region, we first introduce Grounded-SAM to predict the editing region based on the user's textual description, and then combine it with other conditions to perform local editing. To transfer the detail of the given garment texture into the target fashion image, we propose a texture injection and refinement mechanism. Specifically, this mechanism employs a decoupled cross-attention layer to integrate textual descriptions and texture images, and incorporates an auxiliary U-Net to preserve the high-frequency details of generated garment texture. Additionally, we extend the VITON-HD dataset using a multimodal large language model to generate paired samples with texture images and textual descriptions. Extensive experiments show that our DPDEdit outperforms state-of-the-art methods in terms of image fidelity and coherence with the given multimodal inputs.

cs.CV↗

arXiv2025

The Snapshot Problem for the Euler-Poisson-Darboux Equation

The generalized Euler-Poisson-Darboux (EPD) equation with complex parameter $α$ is given by $$ Δ_x u=\frac{\partial^2 u}{\partial t^2}+\frac{n-1+2α}{t}\,\frac{\partial u}{\partial t}, $$ where $u(x,t)\in \mathscr E(\mathbb R^n\times \mathbb R)$, with $u$ even in $t$. For $α=0$ and $α=1$ the solution $u(x,t)$ represents a mean value over spheres and balls, respectively, of radius $|t|$ in $\mathbb R^n$. In this paper we consider existence and uniqueness results for the following two-snapshot problem: for fixed positive real numbers $r$ and $s$ and smooth functions $f$ and $g$ on $\mathbb R^n$, what are the conditions under which there is a solution $u(x,t)$ to the generalized EPD equation such that $u(x,r)=f(x)$ and $u(x,s)=g(x)$? The answer leads to a discovery of Liouville-like numbers related to Bessel functions, and we also study the properties of such numbers.

math.AP↗

arXiv2025

FlexEdit: Marrying Free-Shape Masks to VLLM for Flexible Image Editing

Combining Vision Large Language Models (VLLMs) with diffusion models offers a powerful method for executing image editing tasks based on human language instructions. However, language instructions alone often fall short in accurately conveying user requirements, particularly when users want to add, replace elements in specific areas of an image. Luckily, masks can effectively indicate the exact locations or elements to be edited, while they require users to precisely draw the shapes at the desired locations, which is highly user-unfriendly. To address this, we propose FlexEdit, an end-to-end image editing method that leverages both free-shape masks and language instructions for Flexible Editing. Our approach employs a VLLM in comprehending the image content, mask, and user instructions. Additionally, we introduce the Mask Enhance Adapter (MEA) that fuses the embeddings of the VLLM with the image data, ensuring a seamless integration of mask information and model output embeddings. Furthermore, we construct FSMI-Edit, a benchmark specifically tailored for free-shape mask, including 8 types of free-shape mask. Extensive experiments show that our method achieves state-of-the-art (SOTA) performance in LLM-based image editing, and our simple prompting technique stands out in its effectiveness. The code and data can be found at https://github.com/A-new-b/flex_edit.

cs.CV↗

arXiv2025

Can LLM Improve for Expert Forecast Combination? Evidence from the European Central Bank Survey

This study explores the potential of large language models (LLMs) to enhance expert forecasting through ensemble learning. Leveraging the European Central Bank's Survey of Professional Forecasters (SPF) dataset, we propose a comprehensive framework to evaluate LLM-driven ensemble predictions under varying conditions, including the intensity of expert disagreement, dynamics of herd behavior, and limitations in attention allocation.

stat.AP↗

arXiv2025

A Self-scaled Approximate $\ell_0$ Regularization Robust Model for Outlier Detection

Robust regression models in the presence of outliers have significant practical relevance in areas such as signal processing, financial econometrics, and energy management. Many existing robust regression methods, either grounded in statistical theory or sparse signal recovery, typically rely on the explicit or implicit assumption of outlier sparsity to filter anomalies and recover the underlying signal or data. However, these methods often suffer from limited robustness or high computational complexity, rendering them inefficient for large-scale problems. In this work, we propose a novel robust regression model based on a Self-scaled Approximate l0 Regularization Model (SARM) scheme. By introducing a self-scaling mechanism into the regularization term, the proposed model mitigates the negative impact of uneven or excessively large outlier magnitudes on robustness. We also develop an alternating minimization algorithm grounded in Proximal Operators and Block Coordinate Descent. We rigorously prove the algorithm convergence. Empirical comparisons with several state-of-the-art robust regression methods demonstrate that SARM not only achieves superior robustness but also significantly improves computational efficiency. Motivated by both the theoretical error bound and empirical observations, we further design a Two-Stage SARM (TSSARM) framework, which better utilizes sample information when the singular values of the design matrix are widely spread, thereby enhancing robustness under certain conditions. Finally, we validate our approach on a real-world load forecasting task. The experimental results show that our method substantially enhances the robustness of load forecasting against adversarial data attacks, which is increasingly critical in the era of heightened data security concerns.

eess.SP↗

arXiv2025

Ladder-residual: parallelism-aware architecture for accelerating large model inference with communication overlapping

Large language model inference is both memory-intensive and time-consuming, often requiring distributed algorithms to efficiently scale. Various model parallelism strategies are used in multi-gpu training and inference to partition computation across multiple devices, reducing memory load and computation time. However, using model parallelism necessitates communication of information between GPUs, which has been a major bottleneck and limits the gains obtained by scaling up the number of devices. We introduce Ladder Residual, a simple architectural modification applicable to all residual-based models that enables straightforward overlapping that effectively hides the latency of communication. Our insight is that in addition to systems optimization, one can also redesign the model architecture to decouple communication from computation. While Ladder Residual can allow communication-computation decoupling in conventional parallelism patterns, we focus on Tensor Parallelism in this paper, which is particularly bottlenecked by its heavy communication. For a Transformer model with 70B parameters, applying Ladder Residual to all its layers can achieve 29% end-to-end wall clock speed up at inference time with TP sharding over 8 devices. We refer the resulting Transformer model as the Ladder Transformer. We train a 1B and 3B Ladder Transformer from scratch and observe comparable performance to a standard dense transformer baseline. We also show that it is possible to convert parts of the Llama-3.1 8B model to our Ladder Residual architecture with minimal accuracy degradation by only retraining for 3B tokens. We release our code for training and inference for easier replication of experiments.

cs.LG↗

arXiv2025

TransAug: Translate as Augmentation for Sentence Embeddings

While contrastive learning greatly advances the representation of sentence embeddings, it is still limited by the size of the existing sentence datasets. In this paper, we present TransAug (Translate as Augmentation), which provide the first exploration of utilizing translated sentence pairs as data augmentation for text, and introduce a two-stage paradigm to advances the state-of-the-art sentence embeddings. Instead of adopting an encoder trained in other languages setting, we first distill a Chinese encoder from a SimCSE encoder (pretrained in English), so that their embeddings are close in semantic space, which can be regraded as implicit data augmentation. Then, we only update the English encoder via cross-lingual contrastive learning and frozen the distilled Chinese encoder. Our approach achieves a new state-of-art on standard semantic textual similarity (STS), outperforming both SimCSE and Sentence-T5, and the best performance in corresponding tracks on transfer tasks evaluated by SentEval.

cs.CL↗

arXiv2025

FloE: On-the-Fly MoE Inference on Memory-constrained GPU

With the widespread adoption of Mixture-of-Experts (MoE) models, there is a growing demand for efficient inference on memory-constrained devices. While offloading expert parameters to CPU memory and loading activated experts on demand has emerged as a potential solution, the large size of activated experts overburdens the limited PCIe bandwidth, hindering the effectiveness in latency-sensitive scenarios. To mitigate this, we propose FloE, an on-the-fly MoE inference system on memory-constrained GPUs. FloE is built on the insight that there exists substantial untapped redundancy within sparsely activated experts. It employs various compression techniques on the expert's internal parameter matrices to reduce the data movement load, combined with low-cost sparse prediction, achieving perceptible inference acceleration in wall-clock time on resource-constrained devices. Empirically, FloE achieves a 9.3x compression of parameters per expert in Mixtral-8x7B; enables deployment on a GPU with only 11GB VRAM, reducing the memory footprint by up to 8.5x; and delivers a 48.7x inference speedup compared to DeepSpeed-MII on a single GeForce RTX 3090 - all with only a 4.4$\%$ - 7.6$\%$ average performance degradation.

cs.LG↗

arXiv2025

USTEP: Spatio-Temporal Predictive Learning under A Unified View

Spatio-temporal predictive learning plays a crucial role in self-supervised learning, with wide-ranging applications across a diverse range of fields. Previous approaches for temporal modeling fall into two categories: recurrent-based and recurrent-free methods. The former, while meticulously processing frames one by one, neglect short-term spatio-temporal information redundancies, leading to inefficiencies. The latter naively stack frames sequentially, overlooking the inherent temporal dependencies. In this paper, we re-examine the two dominant temporal modeling approaches within the realm of spatio-temporal predictive learning, offering a unified perspective. Building upon this analysis, we introduce USTEP (Unified Spatio-TEmporal Predictive learning), an innovative framework that reconciles the recurrent-based and recurrent-free methods by integrating both micro-temporal and macro-temporal scales. Extensive experiments on a wide range of spatio-temporal predictive learning demonstrate that USTEP achieves significant improvements over existing temporal modeling approaches, thereby establishing it as a robust solution for a wide range of spatio-temporal applications.

cs.CV↗

arXiv2025

Improving Model Alignment Through Collective Intelligence of Open-Source LLMS

Building helpful and harmless large language models (LLMs) requires effective model alignment approach based on human instructions and feedback, which necessitates high-quality human-labeled data. Constructing such datasets is often expensive and hard to scale, and may face potential limitations on diversity and generalization. To address these challenges, we introduce Mixture of Agents Alignment (MoAA), that leverages the collective strengths of various language models to provide high-quality data for model alignment. By employing MoAA, we enhance both supervised fine-tuning and preference optimization, leading to improved performance compared to using a single model alone to generate alignment data (e.g. using GPT-4o alone). Evaluation results show that our approach can improve win rate of LLaMA-3.1-8B-Instruct from 19.5 to 48.3 on Arena-Hard and from 22.33 to 57.23 on AlpacaEval2, highlighting a promising direction for model alignment through this new scalable and diverse synthetic data recipe. Furthermore, we demonstrate that MoAA enables a self-improvement pipeline, where models finetuned on MoA-generated data surpass their own initial capabilities, providing evidence that our approach can push the frontier of open-source LLMs without reliance on stronger external supervision. Data and code will be released.

cs.CL↗