arXiv subjects

Zhihan Chen

Publications and source records attributed to Zhihan Chen.

12 recordsLinked to original sources

arXiv2026

COMPASS: Grounding Composition-Intent Guidance in Unified Multimodal Models

Composition is a high-level visual intent that governs where subjects are placed and how a scene is organized, yet current unified multimodal models remain unreliable at fine-grained composition recognition and struggle to turn such intent into controllable generation. We present COMPASS, the first unified multimodal framework that grounds composition-intent control in a single system spanning both composition perception and composition-guided generation, with a shared expert token $τ_c$ as the central intent anchor. On the perception side, COMPASS injects composition expertise into an MoE backbone in a minimally invasive manner and distills the inferred intent into $τ_c$. On the generation side, COMPASS reuses $τ_c$ as a global conditioning signal that steers the denoising trajectory, effectively converting passive composition analysis into explicit layout control. To support systematic instruction-following composition learning and evaluation at scale, we construct Comp-11, a large-scale dataset with an 11-class taxonomy and reasoning-augmented annotations. Extensive experiments show that COMPASS substantially improves category-level composition understanding and delivers more composition-consistent, prompt-faithful generation than strong baselines.

cs.AI↗

arXiv2026

MIBE: Multi-subject Interaction Benchmark and Evaluator for Personalized Image Generation

Multi-subject personalized image generation requires the precise rendering of all requested reference identities and their specified interactions based on a guiding prompt. However, state-of-the-art models still struggle with this process, frequently omitting subjects, failing to preserve reference appearances, or misattributing interactions. Furthermore, existing metrics designed primarily for single-subject fidelity cannot reliably capture these errors, suffering severe degradation in ranking separability and failing to align with human preference as the subject count increases. To address this gap, we introduce Multi-subject Interaction Benchmark and Evaluator (MIBE), a unified framework comprising a Multi-subject Interaction Benchmark (MIB) and a Multi-subject Interaction Evaluator (MIE). MIB systematically covers diverse relation types and scene complexities through a decoupled data regime. This consists of a 60K-pair VLM-labeled Silver Set for scalable metric training and a 4K-pair double-blind Human Evaluation Gold Set covering a diverse range of state-of-the-art generators, with the Silver Set reaching 95.1% cross-VLM preference agreement. To demonstrate the utility of this benchmark, we present MIE, a lightweight, reference-conditioned evaluator trained exclusively on the Silver Set with a dual-head ranking and diagnosis objective. MIE exhibits strong cross-generator generalization on the Gold Set, achieving 0.922 overall pairwise accuracy against human preference, including 0.982 on seen generators and 0.884 on unseen generators. By outperforming a broad spectrum of baseline metrics, including CLIP and DINO variants, MIE demonstrates that diagnostic supervision can preserve ranking separability and human alignment where traditional evaluators collapse.

cs.CV↗

arXiv2026

Discovering heuristics in a complex SAT solver with large language models

The Satisfiability problem (SAT) is fundamental in computational complexity theory and has a wide range of industrial applications. Optimizing modern SAT solvers in real-world settings is quite challenging due to their intricate architectures. While automatic configuration frameworks have been developed, they rely on manually constrained search spaces. Here we develop AutoModSAT, a framework that uses large language models (LLMs) to automatically optimize SAT solvers. AutoModSAT combines an LLM-compatible modular solver design, unsupervised prompt optimization to diversify generated functions, and an efficient search procedure based on presearch strategy and a $(1+λ)$ evolutionary algorithm. Extensive experiments across a wide range of datasets demonstrate that AutoModSAT achieves $40\%$ performance improvement over the baseline solver and $30\%$ improvement over the state-of-the-art solvers. Moreover, AutoModSAT also attains a notable speedup compared to the parameter-tuned alternatives of the state-of-the-art solvers over most of the test datasets. These results demonstrate the potential of LLM-guided heuristic discovery for optimizing complex SAT solvers.

cs.AI↗

arXiv2026

When Identities Collapse: A Stress-Test Benchmark for Multi-Subject Personalization

Subject-driven text-to-image diffusion models have achieved remarkable success in preserving single identities, yet their ability to compose multiple interacting subjects remains largely unexplored and highly challenging. Existing evaluation protocols typically rely on global CLIP metrics, which are insensitive to local identity collapse and fail to capture the severity of multi-subject entanglement. In this paper, we identify a pervasive "Illusion of Scalability" in current models: while they excel at synthesizing 2-4 subjects in simple layouts, they suffer from catastrophic identity collapse when scaled to 6-10 subjects or tasked with complex physical interactions. To systematically expose this failure mode, we construct a rigorous stress-test benchmark comprising 75 prompts distributed across varying subject counts and interaction difficulties (Neutral, Occlusion, Interaction). Furthermore, we demonstrate that standard CLIP-based metrics are fundamentally flawed for this task, as they often assign high scores to semantically correct but identity-collapsed images (e.g., generating generic clones). To address this, we introduce the Subject Collapse Rate (SCR), a novel evaluation metric grounded in DINOv2's structural priors, which strictly penalizes local attention leakage and homogenization. Our extensive evaluation of state-of-the-art models (MOSAIC, XVerse, PSR) reveals a precipitous drop in identity fidelity as scene complexity grows, with SCR approaching 100% at 10 subjects. We trace this collapse to the semantic shortcuts inherent in global attention routing, underscoring the urgent need for explicit physical disentanglement in future generative architectures.

cs.CV↗

arXiv2025

FastLEC: Parallel Datapath Equivalence Checking with Hybrid Engines

Combinational equivalence checking (CEC) remains a challenge EDA task in the formal verification of datapath circuits due to their complex arithmetic structures and the limited capability or scalability of SAT, BDD, and exact-simulation (ES) based techniques when used independently. This work presents FastLEC, a hybrid prover that unifies these three formal reasoning engines and introduces three strategies that substantially enhance verification efficiency. First, a regression-based engine-scheduling heuristic predicts solver effectiveness, enabling more accurate and balanced allocation of computational resources. Second, datapath-structure-aware partitioning strategies, along with a dynamic divide-and-conquer SAT prover, exploit the regularity of arithmetic designs while preserving completeness. Third, the memory overhead of ES is significantly reduced through address-reference-count tracking, and simulation is further accelerated through a GPU-enabled backend. FastLEC is evaluated across 368 datapath circuits. Using 32 CPU cores, it proves 5.07x more circuits than the widely used ABC &cec tool. Compared with the latest best datapath-oriented serial and parallel CEC provers, FastLEC outperforms them by 3.33x and 2.67x in PAR-2 time, demonstrating an improvement of 74 newly solved circuits. With the addition of a single GPU, it achieves a further 4.07x improvement. The prover also demonstrates excellent scalability.

cs.LO↗

arXiv2024

Datapath Combinational Equivalence Checking With Hybrid Sweeping Engines and Parallelization

In the application of IC design for microprocessors, there are often demands for optimizing the implementation of datapath circuits, on which various arithmetic operations are performed. Combinational equivalence checking (CEC) plays an essential role in ensuring the correctness of design optimization. The most prevalent CEC algorithms are based on SAT sweeping, which utilizes SAT to prove the equivalence of the internal node pairs in topological order, and the equivalent nodes are merged. Datapath circuits usually contain equivalent pairs for which the transitive fan-in cones are small but have a high XOR chain density, and proving such node pairs is very difficult for SAT solvers. An exact probability-based simulation (EPS) is suitable for verifying such pairs, while this method is not suitable for pairs with many primary inputs due to the memory cost. We first reduce the memory cost of EPS and integrate it to improve the SAT sweeping method. Considering the complementary abilities of SAT and EPS, we design an engine selection heuristic to dynamically choose SAT or EPS in the sweeping process, according to XOR chain density. Our method is further improved by reducing unnecessary engine calls by detecting regularity. Furthermore, we parallelized the SAT and EPS engines of HybridCEC, leading to the parallel CEC prover. Experiments on a benchmark suite from industrial datapath circuits show that our method is much faster than the state-of-the-art CEC tool namely ABC &cec on nearly all instances, and is more than 100x faster on 30% of the instances, 1000x faster on 12% of the instances. In addition, the 64 threads version of our method achieved 77x speedup.

cs.DC↗

arXiv2024

ParLS-PBO: A Parallel Local Search Solver for Pseudo Boolean Optimization

As a broadly applied technique in numerous optimization problems, recently, local search has been employed to solve Pseudo-Boolean Optimization (PBO) problem. A representative local search solver for PBO is LSPBO. In this paper, firstly, we improve LSPBO by a dynamic scoring mechanism, which dynamically strikes a balance between score on hard constraints and score on the objective function. Moreover, on top of this improved LSPBO , we develop the first parallel local search PBO solver. The main idea is to share good solutions among different threads to guide the search, by maintaining a pool of feasible solutions. For evaluating solutions when updating the pool, we propose a function that considers both the solution quality and the diversity of the pool. Furthermore, we calculate the polarity density in the pool to enhance the scoring function of local search. Our empirical experiments show clear benefits of the proposed parallel approach, making it competitive with the parallel version of the famous commercial solver Gurobi.

cs.AI↗

arXiv2024

Revisiting Restarts of CDCL: Should the Search Information be Preserved?

SAT solvers are indispensable in formal verification for hardware and software with many important applications. CDCL is the most widely used framework for modern SAT solvers, and restart is an essential technique of CDCL. When restarting, CDCL solvers cancel the current variable assignment while maintaining the branching order, variable phases, and learnt clauses. This type of restart is referred to as warm restart in this paper. Although different restart policies have been studied, there is no study on whether such information should be kept after restarts. This work addresses this question and finds some interesting observations. This paper indicates that under this popular warm restart scheme, there is a substantial variation in run-time with different randomized initial orders and phases, which motivates us to forget some learned information periodically to prevent being stuck in a disadvantageous search space. We propose a new type of restart called cold restart, which differs from previous restarts by forgetting some of the learned information. Experiments show that modern CDCL solvers can benefit from periodically conducting cold restarts. Based on the analysis of the cold-restart strategies, we develop a parallel SAT solver. Both the sequential and parallel versions of cold restart are more suitable for satisfiable instances, which suggests that existing CDCL heuristics for information management should be revised if one hopes to construct a satisfiable-oriented SAT solver.

cs.LO↗

arXiv2024

Improving Critical Node Detection Using Neural Network-based Initialization in a Genetic Algorithm

The Critical Node Problem (CNP) is concerned with identifying the critical nodes in a complex network. These nodes play a significant role in maintaining the connectivity of the network, and removing them can negatively impact network performance. CNP has been studied extensively due to its numerous real-world applications. Among the different versions of CNP, CNP-1a has gained the most popularity. The primary objective of CNP-1a is to minimize the pair-wise connectivity in the remaining network after deleting a limited number of nodes from a network. Due to the NP-hard nature of CNP-1a, many heuristic/metaheuristic algorithms have been proposed to solve this problem. However, most existing algorithms start with a random initialization, leading to a high cost of obtaining an optimal solution. To improve the efficiency of solving CNP-1a, a knowledge-guided genetic algorithm named K2GA has been proposed. Unlike the standard genetic algorithm framework, K2GA has two main components: a pretrained neural network to obtain prior knowledge on possible critical nodes, and a hybrid genetic algorithm with local search for finding an optimal set of critical nodes based on the knowledge given by the trained neural network. The local search process utilizes a cut node-based greedy strategy. The effectiveness of the proposed knowledgeguided genetic algorithm is verified by experiments on 26 realworld instances of complex networks. Experimental results show that K2GA outperforms the state-of-the-art algorithms regarding the best, median, and average objective values, and improves the best upper bounds on the best objective values for eight realworld instances.

cs.NE↗

arXiv2024

SAR Despeckling via Regional Denoising Diffusion Probabilistic Model

Speckle noise poses a significant challenge in maintaining the quality of synthetic aperture radar (SAR) images, so SAR despeckling techniques have drawn increasing attention. Despite the tremendous advancements of deep learning in fixed-scale SAR image despeckling, these methods still struggle to deal with large-scale SAR images. To address this problem, this paper introduces a novel despeckling approach termed Region Denoising Diffusion Probabilistic Model (R-DDPM) based on generative models. R-DDPM enables versatile despeckling of SAR images across various scales, accomplished within a single training session. Moreover, The artifacts in the fused SAR images can be avoided effectively with the utilization of region-guided inverse sampling. Experiments of our proposed R-DDPM on Sentinel-1 data demonstrates superior performance to existing methods.

cs.CV↗

arXiv2023

Optothermal rotation of micro-/nano-objects in liquids

Controllable rotation of micro-/nano-objects provides tremendous opportunities for cellular biology, three-dimensional (3D) imaging, and micro/nanorobotics. Among different rotation techniques, optical rotation is particularly attractive due to its contactless and fuel-free operation. However, optical rotation precision is typically impaired by the intrinsic optical heating of the target objects. Optothermal rotation, which harnesses light-modulated thermal effects, features simpler optics, lower operational power, and higher applicability to various objects. In this Feature Article, we discuss the recent progress of optothermal rotation with a focus on work from our research group. We categorize the various rotation techniques based on distinct physical mechanisms, including thermophoresis, thermoelectricity, thermo-electrokinetics, thermo-osmosis, thermal convection, and thermo-capillarity. Benefiting from the different rotation modes (i.e., in-plane and out-of-plane rotation), diverse applications in single-cell mechanics, 3D bio-imaging, and micro/nanomotors are demonstrated. We conclude the article with our perspectives on the operating guidelines, existing challenges, and future directions of optothermal rotation.

physics.optics↗

arXiv2021

Hyperspectral Remote Sensing Image Classification Based on Multi-scale Cross Graphic Convolution

The mining and utilization of features directly affect the classification performance of models used in the classification and recognition of hyperspectral remote sensing images. Traditional models usually conduct feature mining from a single perspective, with the features mined being limited and the internal relationships between them being ignored. Consequently, useful features are lost and classification results are unsatisfactory. To fully mine and utilize image features, a new multi-scale feature-mining learning algorithm (MGRNet) is proposed. The model uses principal component analysis to reduce the dimensionality of the original hyperspectral image (HSI) to retain 99.99% of its semantic information and extract dimensionality reduction features. Using a multi-scale convolution algorithm, the input dimensionality reduction features were mined to obtain shallow features, which then served as inputs into a multi-scale graph convolution algorithm to construct the internal relationships between eigenvalues at different scales. We then carried out cross fusion of multi-scale information obtained by graph convolution, before inputting the new information obtained into the residual network algorithm for deep feature mining. Finally, a flexible maximum transfer function classifier was used to predict the final features and complete the classification. Experiments on three common hyperspectral datasets showed the MGRNet algorithm proposed in this paper to be superior to traditional methods in recognition accuracy.

cs.CV↗