arXiv subjects

Kaibo Wang

Publications and source records attributed to Kaibo Wang.

10 recordsLinked to original sources

arXiv2026

OptiMAS: Automatically Optimize Multi-Agent System

Automated evolution of Multi-Agent Systems (MAS) holds significant potential for reducing the manual effort required to design and optimize LLM-based agent architectures. However, extant search-based paradigms face a fundamental trade-off, where an expanded optimization scope exacerbates evolutionary instability, while discrete branch-and-discard search isolates insights across lineages. To address these limitations, we propose a continuous, data-driven optimization paradigm built upon a unified ReAct-based infrastructure that reconciles a broad optimization scope with operational stability. Under this paradigm, we present OptiMAS, a task-agnostic agentic optimizer that leverages textual interaction trajectories and task feedback as loss signals for end-to-end MAS evolution. Equipped with a novel dual-track memory mechanism, OptiMAS sustains performance improvement over extended optimization horizons. Evaluation on four heterogeneous agentic benchmarks with three varying scale and accessibility LLM backbones, demonstrates that OptiMAS consistently achieves competitive or superior accuracy relative to both domain-specialized hand-crafted systems and existing evolutionary methods. Our work establishes a practical milestone toward robust, automated MAS evolution.

cs.MA↗

arXiv2026

Stabilizing Reinforcement Learning for Diffusion Language Models

Group Relative Policy Optimization (GRPO) is highly effective for post-training autoregressive (AR) language models, yet its direct application to diffusion large language models (dLLMs) often triggers reward collapse. We identify two sources of incompatibility. First, GRPO relies on importance ratios defined by sequence probabilities, which are intractable in dLLMs and must be estimated (e.g., via ELBO-based or mean-field likelihood proxies), yielding inherently noisy ratios. Second, standard GRPO's formulation is not designed for estimated ratios: its conditional clipping can be anomalously bypassed by model-agnostic estimation noise, producing gradient spikes, while its fixed group-size normalization amplifies gradient-magnitude fluctuations under high-variance ratio estimates. We show these effects form a self-reinforcing instability loop that drives policy drift and further increases ratio variance. To break this loop, we propose StableDRL, a reformulation of GRPO tailored for dLLMs that uses (i) unconditional clipping to suppress outlier-induced spikes and (ii) self-normalization to constrain updates within the convex hull of per-sample gradients. We further extend StableDRL to block-wise diffusion models via a staircase attention mechanism.

cs.LG↗

arXiv2026

Beyond Masks: Efficient, Flexible Diffusion Language Models via Deletion-Insertion Processes

While Masked Diffusion Language Models (MDLMs) relying on token masking and unmasking have shown promise in language modeling, their computational efficiency and generation flexibility remain constrained by the masking paradigm. In this paper, we propose Deletion-Insertion Diffusion language models (DID) that rigorously formulate token deletion and insertion as discrete diffusion processes, replacing the masking and unmasking processes in current MDLMs. DID improves training and inference efficiency by eliminating two major sources of computational overhead in MDLMs: the computations on non-informative 1) tokens inherent to the paradigm, and 2) tokens introduced in variable-length settings. Furthermore, DID offers greater flexibility by: 1) natively supporting variable-length sequences without requiring fixed-length padding, and 2) an intrinsic self-correction mechanism during generation due to insertion that dynamically adjusts token positions. To train DID, we design a score-based approach that assigns scores to token insertion operations and derive appropriate training objectives. The objectives involve subsequence counting problems, which we efficiently solve via a parallelized dynamic programming algorithm. Our experiments across fixed and variable-length settings demonstrate the advantage of DID over baselines of MDLMs and existing insertion-based LMs, in terms of modeling performance, sampling quality, and training/inference speed, without any hyperparameter tuning.

cs.CL↗

arXiv2026

On the $ε$-Free Inference Complexity of Absorbing Discrete Diffusion

Absorbing discrete diffusion has emerged as a dominant framework for discrete data generation. However, a significant disparity remains between its empirical success and theoretical understanding: existing analyses fail to demonstrate a complexity advantage over the $\mathcal{O}(d \ln(d/ε))$ baseline established for \emph{uniform} discrete diffusion. We bridge this gap by identifying a critical structural advantage: whereas uniform diffusion redundantly re-denoises valid elements, the absorbing scheme denoises each absorbing state exactly once. Leveraging this insight, we introduce \emph{Absorbing-Aware Truncated Uniformization} (AATU). We prove that AATU achieves $ε$-TV convergence with $\mathcal{O}(d \ln d)$ complexity-\emph{independent} of the error tolerance $ε$-thereby strictly outperforming existing uniform baselines. Beyond improving convergence rates, our analysis eliminates the restrictive bounded-score assumption commonly required in prior studies of uniformization-based inference. Furthermore, we extend AATU to time-invariant parameterizations, showing that it naturally adopts an imputation-type inference with a uniformly randomized denoising order. When combined with a lazy update strategy, TV convergence requires only $\mathcal{O}(d)$ discrete score evaluations. These results not only establish a rigorous foundation for absorbing discrete diffusion -- confirming its efficiency in high-accuracy generation -- but also open new avenues for analyzing diffusion-based language models under the masking paradigm.

cs.LG↗

arXiv2026

StablePDENet: Enhancing Neural Operator Stability through Physics-Informed Residual-Sensitivity Regularization

Learning solution operators for differential equations with neural networks has shown great potential in scientific computing, but ensuring their stability under input perturbations remains a critical challenge. We introduce the StablePDENet, a physics-informed adversarial training method that regularizes the residual sensitivity with respect to an input perturbation. The operator learning task is formulated as a min--max optimization problem, where the inner model searches admissible input perturbations by physics-based projected-gradient adversary, while the outer problem combines the attacked physics loss with a normalized residual-sensitivity penalty. Moreover, residual-sensitivity regularization is included to ensure that the local Lipschitz constant of the learned operator is a more accurate approximation to that of the exact operator. We evaluate the StablePDENet on several benchmark problems. Compared with PI-DeepONet and its adversarially trained variant, StablePDENet achieves higher accuracy under adversarial input perturbations while maintaining competitive accuracy on clean inputs. The numerical results also demonstrate that the StablePDENet can effectively improve the generalization accuracy for operator learning. The Helmholtz study further distinguishes learned-model sensitivity from amplification intrinsic to an ill-conditioned solution operator. The results support residual-sensitivity regularization as a practical route to more stable and physically consistent neural PDE operators.

cs.LG↗

arXiv2025

Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations

Classifier-Free Guidance (CFG) is an essential component of text-to-image diffusion models, and understanding and advancing its operational mechanisms remains a central focus of research. Existing approaches stem from divergent theoretical interpretations, thereby limiting the design space and obscuring key design choices. To address this, we propose a unified perspective that reframes conditional guidance as fixed point iterations, seeking to identify a golden path where latents produce consistent outputs under both conditional and unconditional generation. We demonstrate that CFG and its variants constitute a special case of single-step short-interval iteration, which is theoretically proven to exhibit inefficiency. To this end, we introduce Foresight Guidance (FSG), which prioritizes solving longer-interval subproblems in early diffusion stages with increased iterations. Extensive experiments across diverse datasets and model architectures validate the superiority of FSG over state-of-the-art methods in both image quality and computational efficiency. Our work offers novel perspectives for conditional guidance and unlocks the potential of adaptive design.

cs.CV↗

arXiv2025

TweezeEdit: Consistent and Efficient Image Editing with Path Regularization

Large-scale pre-trained diffusion models empower users to edit images through text guidance. However, existing methods often over-align with target prompts while inadequately preserving source image semantics. Such approaches generate target images explicitly or implicitly from the inversion noise of the source images, termed the inversion anchors. We identify this strategy as suboptimal for semantic preservation and inefficient due to elongated editing paths. We propose TweezeEdit, a tuning- and inversion-free framework for consistent and efficient image editing. Our method addresses these limitations by regularizing the entire denoising path rather than relying solely on the inversion anchors, ensuring source semantic retention and shortening editing paths. Guided by gradient-driven regularization, we efficiently inject target prompt semantics along a direct path using a consistency model. Extensive experiments demonstrate TweezeEdit's superior performance in semantic preservation and target alignment, outperforming existing methods. Remarkably, it requires only 12 steps (1.6 seconds per edit), underscoring its potential for real-time applications.

cs.CV↗

arXiv2021

FakeWake: Understanding and Mitigating Fake Wake-up Words of Voice Assistants

In the area of Internet of Things (IoT) voice assistants have become an important interface to operate smart speakers, smartphones, and even automobiles. To save power and protect user privacy, voice assistants send commands to the cloud only if a small set of pre-registered wake-up words are detected. However, voice assistants are shown to be vulnerable to the FakeWake phenomena, whereby they are inadvertently triggered by innocent-sounding fuzzy words. In this paper, we present a systematic investigation of the FakeWake phenomena from three aspects. To start with, we design the first fuzzy word generator to automatically and efficiently produce fuzzy words instead of searching through a swarm of audio materials. We manage to generate 965 fuzzy words covering 8 most popular English and Chinese smart speakers. To explain the causes underlying the FakeWake phenomena, we construct an interpretable tree-based decision model, which reveals phonetic features that contribute to false acceptance of fuzzy words by wake-up word detectors. Finally, we propose remedies to mitigate the effect of FakeWake. The results show that the strengthened models are not only resilient to fuzzy words but also achieve better overall performance on original training datasets.

cs.LG↗

arXiv2021

JST-RR Model: Joint Modeling of Ratings and Reviews in Sentiment-Topic Prediction

Analysis of online reviews has attracted great attention with broad applications. Often times, the textual reviews are coupled with the numerical ratings in the data. In this work, we propose a probabilistic model to accommodate both textual reviews and overall ratings with consideration of their intrinsic connection for a joint sentiment-topic prediction. The key of the proposed method is to develop a unified generative model where the topic modeling is constructed based on review texts and the sentiment prediction is obtained by combining review texts and overall ratings. The inference of model parameters are obtained by an efficient Gibbs sampling procedure. The proposed method can enhance the prediction accuracy of review data and achieve an effective detection of interpretable topics and sentiments. The merits of the proposed method are elaborated by the case study from Amazon datasets and simulation studies.

cs.CL↗

arXiv2012

Accelerating Pathology Image Data Cross-Comparison on CPU-GPU Hybrid Systems

As an important application of spatial databases in pathology imaging analysis, cross-comparing the spatial boundaries of a huge amount of segmented micro-anatomic objects demands extremely data- and compute-intensive operations, requiring high throughput at an affordable cost. However, the performance of spatial database systems has not been satisfactory since their implementations of spatial operations cannot fully utilize the power of modern parallel hardware. In this paper, we provide a customized software solution that exploits GPUs and multi-core CPUs to accelerate spatial cross-comparison in a cost-effective way. Our solution consists of an efficient GPU algorithm and a pipelined system framework with task migration support. Extensive experiments with real-world data sets demonstrate the effectiveness of our solution, which improves the performance of spatial cross-comparison by over 18 times compared with a parallelized spatial database approach.

cs.DB↗