arXiv subjects

Wenzhi Li

Publications and source records attributed to Wenzhi Li.

5 recordsLinked to original sources

arXiv2026

Dual-Layer Agentic Memory with Fast Write Routing and Slow Consolidation

Large language model (LLM) agents operate in dynamic environments where knowledge continuously evolves. Existing memory systems typically treat external memory as a monotonically growing repository, inevitably leading to retrieval degradation and increasing computational costs over time. We argue that the core challenge is not retrieval alone, but managing the knowledge lifecycle: deciding what to externalize, update, or ultimately internalize. Inspired by Complementary Learning Systems (CLS) theory in neuroscience, we propose Dual-Layer Agentic Memory, a framework that shifts memory management to the write phase through cost-aware epistemic routing and periodic parametric consolidation. Incoming information is categorized as non-write, write-new, or write-update, and routed through a small-to-large model cascade that minimizes routing overhead while filtering redundant memories. A subsequent write-back phase selectively consolidates high-value external memories into model parameters via supervised fine-tuning. Experiments demonstrate the dual efficiency of our approach: a 1.7B/8B cascade prunes up to 68% of redundant external memory while escalating fewer than 50% of inputs, yet retains over 98% of the downstream QA Exact Match (EM) achieved by an exhaustive retention baseline. We further show that periodic consolidation successfully internalizes external knowledge, allowing the router to adaptively suppress redundant writes as the model's epistemic boundaries evolve. Overall, our framework presents a unified paradigm for agent memory: selective externalization followed by selective internalization. Code and dataset will be released upon acceptance.

cs.CL↗

arXiv2026

ChainPrune: Evaluating and Reducing Redundancy in Long Chain-of-Thought Reasoning

Chain-of-Thought (CoT) reasoning has significantly enhanced the multi-step problem-solving capabilities of large language models (LLMs) by introducing explicit intermediate reasoning. However, advanced Large Reasoning Models (LRMs) often exhibit overthinking behaviors, including excessively long reasoning steps, redundant steps, and high computational overhead. Existing token-length reward strategies aim to promote concise outputs, but often result in pseudo-conciseness, where token count is reduced, yet redundant reasoning persists, leading to longer and less structurally efficient chains. To address these limitations, we propose ChainPrune, a novel reasoning path semantic structural optimization method to efficiently and controllably synthesize self-generated high-quality training data. We initially consolidate self-generated reasoning paths into a tree-based structure, followed by a multi-criteria dominant path selection process for preference data construction that formulates shallow reasoning trajectories while preserving essential reasoning steps. To further enhance the quality of reasoning, we incorporate a DPO-based preference learning method combined with supervised loss, effectively mitigating false reward suppression. This innovative integration significantly enhances both the efficiency and effectiveness of our reasoning framework. Comprehensive experimental results demonstrate significant reductions in step length and computational overhead, while maintaining or even enhancing accuracy.

cs.LG↗

arXiv2026

Geometry-Aware Resource Allocation for Network-Level ISAC Systems

Network-level integrated sensing and communication (ISAC) is recognized as a transformative technology for next-generation mobile radio systems. By enabling collaboration among multiple transceivers, network-level ISAC can significantly enhance both communication and sensing performance through spatial diversity. However, existing resource allocation strategies typically overlook the impact of spatial geometry, where identical time-frequency resources contribute differently to sensing accuracy depending on the transceiver's location. This leaves the fundamental coupling between spatial topology and resource efficacy unclear, rendering optimal resource allocation a critical challenge for unlocking the full potential of network-level ISAC.To address this challenge, this paper investigates the optimal distribution of time-frequency resources across spatially distributed transceivers through a theoretically grounded two-stage framework. First, we analytically derive the optimal time and frequency aperture distributions for sensing, defined as the variances of the allocated symbol and subcarrier indices, respectively, under both two-transmitter and multi-transmitter scenarios. By exploiting the mathematical isomorphism between delay and Doppler estimation, we prove that the optimal resource allocation strategy follows the gradient direction of the Cramer-Rao Lower Bound (CRLB) with respect to the apertures. Second, to bridge the gap between theoretical aperture values and practical OFDMA constraints, such as the minimized communication rate of each user equipment (UE), we formulate the resource allocation as a combinatorial integer partitioning problem. To tackle the NP-hard nature of the formulated problem, a low-complexity Variance-Guided Partitioning Algorithm (VGPA) is proposed to jointly optimize the subcarrier and symbol patterns for communication and sensing.

eess.SP↗

arXiv2025

Cat-Eye Inspired Active-Passive-Composite Aperture-Shared Sub-Terahertz Meta-Imager for Non-Interactive Concealed Object Detection

Within the feline eye, a distinctive tapetum lucidum as a mirror resides posterior to the retina, reflecting the incident rays to simulate light source emission. This secondary emission property enables felines to be highly sensitive to light, possessing remarkable visual capabilities even in dark settings. Drawing inspiration from this natural phenomenon, we propose an active-passive-composite sub-terahertz meta-imager integrated with a bifocus metasurface, a high-sensitivity radiometer, and a low-power signal hidden radiation source. Benefiting from its aperture-shared advantage, this advanced fusion imaging system, enabled to be deployed by a simplified portable hardware platform, allows for the concurrent acquisition of active and passive electromagnetic properties to extend the target detection category and realize multi-mode fusion perception. Notably, it also enables the extraction of radiation and reflection characteristics without additional calibration modules. Experiments demonstrate the multi-target fusion imaging and localized information decoupling with the tailored field of view and emission energy. This compact and multi-mode fusion imaging system may have plenty of potential for airplane navigation positioning, abnormal monitoring, and non-interactive concealed security checks.

physics.optics↗

arXiv2013

A Matrix-Field Weighted Mean-Square-Error Model for MIMO Transceiver Designs

In this letter, we investigate an important and famous issue, namely weighted mean-square-error (MSE) minimization transceiver designs. In our work, for transceiver designs a novel weighted MSE model is proposed, which is defined as a linear matrix function with respect to the traditional data detection MSE matrix. The new model can be interpreted an extension of weighting operation from vector field to matrix field. Based on the proposed weighting operation a general transceiver design is proposed, which aims at minimizing an increasing matrix-monotone function of the output of the previous linear matrix function. The structure of the optimal solutions is also derived. Furthermore, two important special cases of the matrix-monotone functions are discussed in detail. It is also revealed that these two problems are exactly equivalent to the transceiver designs of sum MSE minimization and capacity maximization for dual-hop amplify-and-forward (AF) MIMO relaying systems, respectively. Finally, it is concluded that the AF relaying is undoubtedly this kind of weighting operation.

cs.IT↗