Recent advancements in Large Language Models (LLMs) have shown that it is promising to utilize Process Reward Models (PRMs) as verifiers to enhance the performance of LLMs. However, current PRMs face three key challenges: (1) limited process supervision and generalization capabilities, (2) dependence on scalar value prediction without leveraging the generative abilities of LLMs, and (3) inability to scale the test-time compute of PRMs. In this work, we introduce GenPRM, a generative process reward model that performs explicit Chain-of-Thought (CoT) reasoning with code verification before providing judgment for each reasoning step. To obtain high-quality process supervision labels and rationale data, we propose Relative Progress Estimation (RPE) and a rationale synthesis framework that incorporates code verification. Experimental results on ProcessBench and several mathematical reasoning tasks show that GenPRM significantly outperforms prior PRMs with only 23K training data from MATH dataset. Through test-time scaling, a 1.5B GenPRM outperforms GPT-4o, and a 7B GenPRM surpasses Qwen2.5-Math-PRM-72B on ProcessBench. Additionally, GenPRM demonstrates strong abilities to serve as a critic model for policy model refinement. This work establishes a new paradigm for process supervision that bridges the gap between PRMs and critic models in LLMs.

Fangyuan Li, Pengfei Li, Shijie Wang, Junqi Gao, Jianxing Liu, Biqing Qi, Yuqiang Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Binary Neural Networks (BNNs) have garnered significant attention due to their immense potential for deployment on edge devices. However, the non-differentiability of the quantization function poses a challenge for the optimization of BNNs, as its derivative cannot be backpropagated. To address this issue, hypernetwork based methods, which utilize neural networks to learn the gradients of non-differentiable quantization functions, have emerged as a promising approach due to their adaptive learning capabilities to reduce estimation errors. However, existing hypernetwork based methods typically rely solely on current gradient information, neglecting the influence of historical gradients. This oversight can lead to accumulated gradient errors when calculating gradient momentum during optimization. To incorporate historical gradient information, we design a Historical Gradient Storage (HGS) module, which models the historical gradient sequence to generate the first-order momentum required for optimization. To further enhance gradient generation in hypernetworks, we propose a Fast and Slow Gradient Generation (FSG) method. Additionally, to produce more precise gradients, we introduce Layer Recognition Embeddings (LRE) into the hypernetwork, facilitating the generation of layer-specific fine gradients. Extensive comparative experiments on the CIFAR-10 and CIFAR-100 datasets demonstrate that our method achieves faster convergence and lower loss values, outperforming existing baselines.

Graph Retrieval Augmented Generation (GraphRAG) effectively enhances external knowledge integration capabilities by explicitly modeling knowledge relationships, thereby improving the factual accuracy and generation quality of Large Language Models (LLMs) in specialized domains.However, existing methods suffer from two inherent limitations: 1) Inefficient Information Aggregation: They rely on a single agent and fixed iterative patterns, making it difficult to adaptively capture multi-level textual, structural, and degree information within graph data.2) Rigid Reasoning Mechanism: They employ preset reasoning schemes, which cannot dynamically adjust reasoning depth nor achieve precise semantic correction.To overcome these limitations, we propose Graph Counselor, an GraphRAG method based on multi-agent collaboration.This method uses the Adaptive Graph Information Extraction Module (AGIEM), where Planning, Thought, and Execution Agents work together to precisely model complex graph structures and dynamically adjust information extraction strategies, addressing the challenges of multi-level dependency modeling and adaptive reasoning depth.Additionally, the Self-Reflection with Multiple Perspectives (SR) module improves the accuracy and semantic consistency of reasoning results through self-reflection and backward reasoning mechanisms.Experiments demonstrate that Graph Counselor outperforms existing methods in multiple graph reasoning tasks, exhibiting higher reasoning accuracy and generalization ability.Our code is available at Graph-Counselor.

As an effective approach to equip models with multitask capabilities without additional training, model merging has garnered significant attention. However, existing merging methods face challenges of redundant parameter conflicts and the excessive storage burden of fine-tuned parameters. In this work, through controlled experiments, we reveal that for fine-tuned task vectors, only those parameters with magnitudes above a certain threshold contribute positively to the task, exhibiting a pulse-like characteristic. We then attempt leveraging this pulse-like characteristic to binarize the task vectors and reduce storage overhead. Further controlled experiments show that the binarized task vectors incur almost no decrease in fine-tuning and merging performance, and even exhibit stronger performance improvements as the proportion of redundant parameters increases. Based on these insights, we propose Task Switch (T-Switch), which decomposes task vectors into three components: 1) an activation switch instantiated by a binarized mask vector, 2) a polarity switch instantiated by a binarized sign vector, and 3) a scaling knob instantiated by a scalar coefficient. By storing task vectors in a binarized form, T-Switch alleviates parameter conflicts while ensuring efficient task parameter storage. Furthermore, to enable automated switch combination in T-Switch, we further introduce Auto-Switch, which enables training-free switch combination via retrieval from a small query set. Experiments indicate that our methods achieve significant performance improvements over existing baselines, requiring only 1-3% of the storage space of full-precision parameters.

Sihang Zeng, Kai Tian, Kaiyan Zhang, Yuru Wang, Junqi Gao, Runze Liu, Sa Yang, Jingxuan Li, Xinwei Long, Jiaheng Ma, Biqing Qi, Bowen Zhou. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Heterogeneous Large Language Model (LLM) fusion integrates the strengths of multiple source LLMs with different architectures into a target LLM with low computational overhead. While promising, existing methods suffer from two major limitations: 1) **reliance on real data from limited domain** for knowledge fusion, preventing the target LLM from fully acquiring knowledge across diverse domains, and 2) **fixed data allocation proportions** across domains, failing to dynamically adjust according to the target LLM's varying capabilities across domains, leading to a capability imbalance. To overcome these limitations, we propose Bohdi, a synthetic-data-only heterogeneous LLM fusion framework. Through the organization of knowledge domains into a hierarchical tree structure, Bohdi enables automatic domain exploration and multi-domain data generation through multi-model collaboration, thereby comprehensively extracting knowledge from source LLMs. By formalizing domain expansion and data sampling proportion allocation on the knowledge tree as a Hierarchical Multi-Armed Bandit problem, Bohdi leverages the designed DynaBranches mechanism to adaptively adjust sampling proportions based on the target LLM's performance feedback across domains. Integrated with our proposed Introspection-Rebirth (IR) mechanism, DynaBranches dynamically tracks capability shifts during target LLM's updates via Sliding Window Binomial Likelihood Ratio Testing (SWBLRT), further enhancing its online adaptation capability. Comparative experimental results on a comprehensive suite of benchmarks demonstrate that Bohdi significantly outperforms existing baselines on multiple target LLMs, exhibits higher data efficiency, and virtually eliminates the imbalance in the target LLM's capabilities.

Utilizing a rotor drone equipped with a magnetometer presents a practical solution for rapidly surveying unexploded ordance (UXO) areas potentially buried underground. This method effectively reduces operational risks and enhances overall efficiency. Our study has developed an aeromagnetic gradient detection system featuring a scalar sensor mounted on a rotor-based unmanned aerial vehicle(UAV). Experimental results affirm the effectiveness of differential processing in mitigating common-mode noise from the carrier platform and in neutralizing the impact of geomagnetic gradients. The system's dynamic noise is Si=0.013 nT. The wavelet entropy reduction algorithm, utilizing a 'sym6' wavelet basis with j=4 wavelet decomposition, increases the low signal-to-noise ratio SNR= -0.78 dB to 8.46 dB, thereby improving the detection of weak magnetic signals. The system's ultimate depth detection capability for eight buried unexploded bombs has been quantified: the H30, with a length of 17 cm and a mass of 0.3 kg, has a maximum detection range of H=120 cm; the 120 mm caliber projectile, with a length of 240 cm and a mass of 25.8 kg, achieves a maximum detection range of H=550 cm. Such detection capability meets practical detection needs. Furthermore, we perform 50 test sets to assess its detection performance, with four types UXOs randomly placed within a 450 m2area. It took 270 seconds to scan the 450 m2 area and 60 seconds to process data . The system demonstrates a correct detection rate of 94.5% while maintaining a false alarm rate of 2%.

Few-Shot Class-Incremental Learning (FSCIL) has gained considerable attention in recent years for its pivotal role in addressing continuously arriving classes. However, it encounters additional challenges. The scarcity of samples in new sessions intensifies overfitting, causing incompatibility between the output features of new and old classes, thereby escalating catastrophic forgetting. A prevalent strategy involves mitigating catastrophic forgetting through the Explicit Memory (EM), which comprise of class prototypes. However, current EM-based methods retrieves memory globally by performing Vector-to-Vector (V2V) interaction between features corresponding to the input and prototypes stored in EM, neglecting the geometric structure of local features. This hinders the accurate modeling of their positional relationships. To incorporate information of local geometric structure, we extend the V2V interaction to Graph-to-Graph (G2G) interaction. For enhancing local structures for better G2G alignment and the prevention of local feature collapse, we propose the Local Graph Preservation (LGP) mechanism. Additionally, to address sample scarcity in classes from new sessions, the Contrast-Augmented G2G (CAG2G) is introduced to promote the aggregation of same class features thus helps few-shot learning. Extensive comparisons on CIFAR100, CUB200, and the challenging ImageNet-R dataset demonstrate the superiority of our method over existing methods.

A vertical gradiometer is constructed using two scalar magnetometers, with the vertical scalar gradient derived through the application of the Anderson function. The Particle Swarm Optimization (PSO) algorithm is then employed to estimate the target’s position, magnetic moment, and depth by integrating the detection trajectory, magnetic field, and vertical gradient. By thoroughly investigating the impact of computation timetcand particle numberNon algorithm precision, we propose an optimal configuration framework based on Monte Carlo simulation for achieving optimal precision attc=40s andN=100. This algorithm is validated via experiments conducted with a handheld detection system over a 96 m² area, where three types of targets are randomly placed. A comparative analysis of parameter inversion accuracy among the Genetic algorithm (GA), Least Squares Method (LSM), and PSO algorithms reveals that the PSO algorithm achieves the highest accuracy, with maximum spatial errore(x, y, z) max= (0.10 m, 0.13 m, 0.10 m). Specifically, the vetical error isezmax≤ 0.10 m and the average magnetic moment estimation error iseave= 5.23%.

Efficient detection of near-seafloor metallic targets poses significant challenges for conventional underwater inspection methods. In this paper, the optimal detection frequency of the marine electromagnetic detection technology for detecting small targets such as unexploded ordnance (UXO), iron pipes, and aluminum pipes on the seabed is simulated and analyzed using the Monte Carlo method. This analysis is based on the electromagnetic response law of the target and the attenuation law of the electromagnetic field in the seawater medium in marine electromagnetic detection technology. The optimal detection frequency band is obtained to be [1807, 5365] Hz. Selecting a single frequency point in this frequency band to construct a series resonant circuit can provide the optimal radiant efficiency of the transmitter. On this basis, a remotely operated vehicle (ROV)-based magnetic detection system was designed, using a coil as the transmitter and a fluxgate as the receiver. The system was verified through sea trials, and the detection distance of small aluminum pipe targets can reach over 1.5 m.

In this study, an Archimedean Spiral Aeromagnetic (ASA) method is proposed to improve the searching efficiency by unmanned aerial vehicles (UAVs) for magnetic detection. Based on the Monte Carlo method, five thousand searching experiments are simulated. The results show a 16.7 % reduction in average searching time by spiral trajectory (ST) searching compared with the traditional comb trajectory (CT) flight planning.PST, the average probability of faster detection by ST, is 21.0 % higher than that of comb-type PCT for small targets search. For large targets detection,PSTis 22.1 % higher than that ofPCT. Then, the outdoor aeromagnetic detection experiments are conducted in an area of 900 m2and 250,000 m2to search for an unexplored ordnance and a ship, respectively. The results demonstrate that ST can significantly improve the searching efficiency.

Graph Transformers have garnered significant attention due to their ability to address the challenges of long-distance interactions in previous GNNs. However, most current graph Transformers face difficulties when dealing with heterophilic graphs. To investigate this issue, we first analyzed the distribution of attention weights for homophilic and heterophilic graphs. We discovered that heterophily interferes with the allocation of attention weights, leading to errors in node classification. Further investigation revealed that the root cause may be the difficulty of current graph Transformers in capturing the difference between the features of each node and its neighbors. To alleviate this issue, we propose a position encoding strategy called DiSP to better capture the feature difference and introduce FDphormer, a new efficient and simple graph Transformer model based on DiSP. Additionally, we analyze the generalization error of existing graph Transformer models and provide an upper bound on the generalization error of current graph Transformers with the introduction of DiSP. Extensive experiments demonstrate that FDphormer not only outperforms state-of-the-art methods on diverse heterogeneous datasets but also exhibits competitive performance under homophily.

Advanced life forms, sustained by the synergistic interaction of neural cognitive mechanisms, continually acquire and transfer knowledge throughout their lifespan. In contrast, contemporary machine learning paradigms exhibit limitations in emulating the facets of continual learning (CL). Nonetheless, the emergence of large language models (LLMs) presents promising avenues for realizing CL via interactions with these models. Drawing on Complementary Learning System theory, this paper presents a novel Interactive Continual Learning (ICL) framework, enabled by collaborative interactions among models of various sizes. Specifically, we assign the ViT model as System1 and multimodal LLM as System2. To enable the memory module to deduce tasks from class information and enhance Set2Set retrieval, we propose the Class-Knowledge-Task Multi-Head Attention (CKT-MHA). Additionally, to improve memory retrieval in System1 through enhanced geometric representation, we introduce the CL-vMF mechanism, based on the von Mises-Fisher (vMF) distribution. Mean-while, we introduce the von Mises-Fisher Outlier Detection and Interaction (vMF-ODI) strategy to identify hard examples, thus enhancing collaboration between System1 and System2 for complex reasoning realization. Comprehensive evaluation of our proposed ICL demonstrates significant resistance to forgetting and superior performance relative to existing methods. Code is available at github.com/ICL.

Graph incremental learning has gained widespread attention for its ability to mitigate catastrophic forgetting for graph neural networks (GNN). Conventional methods typically require numerous labels for node classification. However, obtaining abundant labels is often challenging in practice, which makes graph few-shot incremental learning necessary. Current approaches rely on large number of samples from meta-learning to construct memories, and heavy fine-tuning of the GNN parameters that lead to the loss of past knowledge. These result in significant memory consumption and loss of past knowledge information, respectively. To tackle these issues, We introduce Mecoin to efficient construct and Preserve memory. For efficient storage and update of class prototypes, Mecoin use Structured Memory Unit (SMU) to cache prototypes of the seen classes and update new class prototypes through interaction between nodes and the cached prototypes by Memory Construction module(MeCo). Besides, to avoid extensive parameter fine-tuning and forgetting, we introduce a Memory Representation Adaptive Module called MRaM to separate the learning of prototypes and class representations and use Graph Knowledge Interchange Module (GKIM) to injects past knowledge information into GNN. We analyze the effectiveness of our paradigm from the perspectives of generalization error, and discuss the impact of different distillation methods on model performance through experiments and VC-dimension. By comparison with other related methods, we validate that Mecoin achieves higher accuracy and lower forgetting rate.

Deep State Space Models (SSMs) have proven effective in numerous task scenarios but face significant security challenges due to Adversarial Perturbations (APs) in real-world deployments. Adversarial Training (AT) is a mainstream approach to enhancing Adversarial Robustness (AR) and has been validated on various traditional DNN architectures. However, its effectiveness in improving the AR of SSMs remains unclear. While many enhancements in SSM components, such as integrating Attention mechanisms and expanding to data-dependent SSM parameterizations, have brought significant gains in Standard Training (ST) settings, their potential benefits in AT remain unexplored. To investigate this, we evaluate existing structural variants of SSMs with AT to assess their AR performance. We observe that pure SSM structures struggle to benefit from AT, whereas incorporating Attention yields a markedly better trade-off between robustness and generalization for SSMs in AT compared to other components. Nonetheless, the integration of Attention also leads to Robust Overfitting (RO) issues. To understand these phenomena, we empirically and theoretically analyze the output error of SSMs under AP. We find that fixed-parameterized SSMs have output error bounds strictly related to their parameters, limiting their AT benefits, while input-dependent SSMs may face the problem of error explosion. Furthermore, we show that the Attention component effectively scales the output error of SSMs during training, enabling them to benefit more from AT, but at the cost of introducing RO due to its high model complexity. Inspired by this, we propose a simple and effective Adaptive Scaling (AdS) mechanism that brings AT performance close to Attention-integrated SSMs without introducing the issue of RO.

From the perspective of information bottleneck (IB) theory, we propose a novel framework for performing black-box transferable adversarial attacks named IBTA, which leverages advancements in invariant features. Intuitively, diminishing the reliance of adversarial perturbations on the original data, under equivalent attack performance constraints, encourages a greater reliance on invariant features that contributes most to classification, thereby enhancing the transferability of adversarial attacks. Building on this motivation, we redefine the optimization of transferable attacks using a novel theoretical framework that centers around IB. Specifically, to overcome the challenge of unoptimizable mutual information, we propose a simple and efficient mutual information lower bound (MILB) for approximating computation. Moreover, to quantitatively evaluate mutual information, we utilize the Mutual Information Neural Estimator (MINE) to perform a thorough analysis. Our experiments on the ImageNet dataset well demonstrate the efficiency and scalability of IBTA and derived MILB. Our code is available at github.com/IBTA.

The transferability of adversarial perturbations provides an effective shortcut for black-box attacks. Targeted perturbations have greater practicality but are more difficult to transfer between models. In this paper, we experimentally and theoretically demonstrated that neural networks trained on the same dataset have more consistent performance in High-Sample-Density-Regions (HSDR) of each class instead of low sample density regions. Therefore, in the target setting, adding perturbations towards HSDR of the target class is more effective in improving transferability. However, density estimation is challenging in high-dimensional scenarios. Further theoretical and experimental verification demonstrates that easy samples with low loss are more likely to be located in HSDR. Perturbations towards such easy samples in the target class can avoid density estimation for HSDR location. Based on the above facts, we verified that adding perturbations to easy samples in the target class improves targeted adversarial transferability of existing attack methods. A generative targeted attack strategy named Easy Sample Matching Attack (ESMA) is proposed, which has a higher success rate for targeted attacks and outperforms the SOTA generative method. Moreover, ESMA requires only $5\%$ of the storage space and much less computation time comparing to the current SOTA, as ESMA attacks all classes with only one model instead of seperate models for each class. Our code is available at https://github.com/gjq100/ESMA

A novel vehicle localization and tracking methods are presented based on magnetic anomaly detection by distributed magnetic sensors. First, taking advantage of total magnetic field, in this article, we propose a total field matching (TFM) method that is immune of rotational vibrations to perform target localization. Instead of directly inverting the nonlinear magnetic dipole equations, we use the TFM approach to find the suboptimal target position, and then apply the linear Kalman filter to tail after the target. Because the relationship is linear between the target dynamics and the localization equations. A case study is performed by simulation to result in an estimated trajectory of (d,ϕ) = (70.8 m, 44.9°) that agrees well with the real one of (d,ϕ) = (70.5 m, 45°). For a vehicle tracking, the outdoors experiment results show good estimation accuracy based on four different sensor networking configurations.

This letter reports a noise suppression method for vector magnetic anomaly detection (MAD) based on noise spatial characteristics analysis. The environment noise indicates space anisotropy that is essential for noise depression. To find a projection direction$e$that allows the projected noise, denoted as$N_{e} = e\cdot N$, to have minimum fluctuations, we compute the standard deviation$\sigma $of$N_{e}$. The projected signal is then processed by a vector orthogonal basis functions (OBFs) detector resulting in a higher signal to noise ratio (SNR) by 3 dB than the traditional OBFs filter. After noise space transformation (NST) processing, the compelling low-frequency noise is suppressed and the weak target signal can be obtained.

This letter presents a frequency characteristics analysis for a magnetic anomaly detection (MAD) study. A magnetic dipole model is developed to study the energy distribution of the obtained MAD signal in the frequency domain, where$f_{h}$is proposed to define the signal high-frequency boundary. The relations between$f_{h}$and other parameters, such as the velocity of the target (or platform), closest path approach (CPA), magnetic moment strength, magnetic moment orientation (MMO), and sensor orientation, are analyzed. We find that the frequency response is independent of the moment strength. The combined effects of the MMO and sensor orientation on$f_{h}$are defined by the functio$n g(\alpha $,$\beta $,$\theta $,$\phi$), which is studied by the Monte Carlo method, yielding a maximum value of max($g) \approx ~0.85$. Thus, the function$f_{h} = 0.85v$/CPA is advanced for frequency boundary estimation. The proposed theoretical analysis is examined in an experiment to aid in extracting the magnetic anomaly signal effectively.