arXiv subjects

Lei Li

Publications and source records attributed to Lei Li.

At least 127 records · Page 7Linked to original sources

arXiv2026

AnalogRetriever: Learning Cross-Modal Representations for Analog Circuit Retrieval

Analog circuit design relies heavily on reusing existing intellectual property (IP), yet searching across heterogeneous representations such as SPICE netlists, schematics, and functional descriptions remains challenging. Existing methods are largely limited to exact matching within a single modality, failing to capture cross-modal semantic relationships. To bridge this gap, we present AnalogRetriever, a unified tri-modal retrieval framework for analog circuit search. We first build a high-quality dataset on top of Masala-CHAI through a two-stage repair pipeline that raises the netlist compile rate from 22\% to 100\%. Built on this foundation, AnalogRetriever encodes schematics and descriptions with a vision-language model and netlists with a port-aware relational graph convolutional network, mapping all three modalities into a shared embedding space via curriculum contrastive learning. Experiments show that AnalogRetriever achieves an average Recall@1 of 75.2\% across all six cross-modal retrieval directions, significantly outperforming existing baselines. When integrated into the AnalogCoder agentic framework as a retrieval-augmented generation module, it consistently improves functional pass rates and enables previously unsolved tasks to be completed. Our code and dataset will be released.

cs.CV↗

arXiv2026

Unlock the Potential of Large Language Models for Predictive Tabular Tasks in Data Science with Table-Specific Pretraining

In the domain of data science, the predictive tasks of classification, regression, and imputation of missing values are commonly encountered challenges associated with tabular data. This research endeavors to apply Large Language Models (LLMs) towards addressing these predictive tasks. Despite their proficiency in comprehending natural language, LLMs fall short in dealing with structured tabular data. This limitation stems from their lacking exposure to the intricacies of tabular data during their foundational training. Our research aims to mitigate this gap by compiling a comprehensive corpus of tables annotated with instructions and executing large-scale training of Llama-2 on this enriched dataset. Furthermore, we investigate the practical application of applying the trained model to zero-shot prediction, few-shot prediction, and in-context learning scenarios. Through extensive experiments, our methodology has shown significant improvements over existing benchmarks. These advancements highlight the efficacy of tailoring LLM training to solve table-related problems in data science, thereby establishing a new benchmark in the utilization of LLMs for enhancing tabular intelligence.

cs.LG↗

arXiv2026

Hierarchical Policy Optimization for Simultaneous Translation of Unbounded Speech

Simultaneous speech translation (SST) generates translations while receiving partial speech input. Recent advances show that large language models (LLMs) can substantially improve SST quality, but at the cost of high computational overhead. To reduce this cost, prior work reformulates SST as a multi-turn dialogue task, enabling full reuse of the LLM's key-value (KV) cache and eliminating redundant feature recomputation. However, this approach relies on supervised fine-tuning (SFT) data in dialogue form, for which few human annotations exist, and existing synthesis methods cannot guarantee data quality. In this work, we propose a Hierarchical Policy Optimization (HPO) approach that post-train models trained on imperfect SFT data. We introduce a hierarchical reward that balances translation quality and latency objectives. Experiments on English to Chinese/German/Japanese demonstrate improvements of over +7 COMET score and +1.25 MetricX score at a latency of 1.5 seconds. Comprehensive ablation studies further validate the effectiveness of different quality rewards, hierarchical reward formulations, and segmentation strategies. Code can be found here https://github.com/owaski/HPO

cs.CL↗

arXiv2026

QoS-Constrained Scheduling in Multi-Cell Multi-User MIMO Networks

In 5G and beyond networks, efficient scheduling is essential to exploit the gains of multi-user MIMO (MU-MIMO) equipped with carrier aggregation and joint transmission (JT). However, cross-cell and cross-carrier scheduling under QoS constraints is challenging due to the strong coupling across users, base stations, and carriers. In this work, we address this problem in multi-cell MU-MIMO networks to maximize system throughput for both JT and non-JT users under rate constraints. The optimization is highly complex as scheduling variables and beamforming (BF) vectors are intertwined. To tackle it, we propose an approximate but tractable surrogate by leveraging the eigen-based zero-forcing BF and massive MIMO asymptotics. The reformulated problem has a separable structure and is amenable to efficient solutions by a penalty-based block coordinate descent method. Simulations demonstrate that the proposed scheduler not only meets the QoS requirements well but also achieves remarkable throughput gains over existing schemes.

eess.SP↗

arXiv2026

Inflated Excellence or True Performance? Rethinking Medical Diagnostic Benchmarks with Dynamic Evaluation

Medical diagnostics is a high-stakes and complex domain that is critical to patient care. However, current evaluations of large language models (LLMs) remain limited in capturing key challenges of clinical diagnostic scenarios. Most rely on benchmarks derived from public exams, raising contamination bias that can inflate performance, and they overlook the confounded nature of real consultations beyond textbook cases. Recent dynamic evaluations offer a promising alternative, but often remain insufficient for diagnosis-oriented benchmarking, with limited coverage of clinically grounded confounders and trustworthiness beyond accuracy. To address these gaps, we propose DyReMe, a dynamic benchmark for medical diagnostics that provides a controlled and scalable stress test of diagnostic robustness. Unlike static exam-style questions, DyReMe generates fresh, consultation-style cases that incorporate clinically grounded confounders, such as differential diagnoses and common misdiagnosis factors. It also varies expression styles to capture heterogeneous patient-style descriptions. Beyond accuracy, DyReMe evaluates LLMs on three additional clinically relevant dimensions: veracity, helpfulness, and consistency. Our experiments show that this dynamic approach yields more challenging assessments and exposes substantial weaknesses of stateof-the-art LLMs under clinically confounded diagnostic settings. These findings highlight the urgent need for evaluation frameworks that better assess trustworthy medical diagnostics 1 under clinically grounded confounders.

cs.CL↗

arXiv2026

Bridging the Gap between User Intent and LLM: A Requirement Alignment Approach for Code Generation

Code generation refers to automatically producing executable programs from user requirements. Recently, researchers have explored approaches to enhance the correctness of generated code with advanced large language models. Although achieving improvements, existing approaches focus on designing reasoning strategies or post-refinement methods to enhance code generation performance. Despite their differences, all these methods share a common assumption: the LLM can correctly understand the given requirement. However, this assumption does not always hold. To fill this gap, we propose REA-Coder, a requirement alignment approach to enhance the code generation performance of LLMs. REA-Coder involves first identifying the requirement content that does not align with LLMs and aligning the requirements. Then, based on the aligned requirements, LLMs generate code and further verify whether the generated code aligns with the requirements, iterating this process of requirement alignment and code generation until generating correct code or achieving the maximum number of iterations. Experimental results show that REA-Coder outperforms all advanced baselines on four LLMs across five programming benchmarks. Concretely, REA-Coder achieves average improvements of 7.93%, 30.25%, 26.75%, 8.59%, and 8.64% on the five benchmark datasets, demonstrating the effectiveness of requirement alignment for improving the code generation performance of LLMs.

cs.SE↗

arXiv2026

AdaSplash-2: Faster Differentiable Sparse Attention

Sparse attention has been proposed as a way to alleviate the quadratic cost of transformers, a central bottleneck in long-context training. A promising line of work is $α$-entmax attention, a differentiable sparse alternative to softmax that enables input-dependent sparsity yet has lagged behind softmax due to the computational overhead necessary to compute the normalizer $τ$. In this paper, we introduce AdaSplash-2, which addresses this limitation through a novel histogram-based initialization that reduces the number of iterations needed to compute $τ$ to typically 1--2. The key idea is to compute a coarse histogram of attention scores on the fly and store it in on-chip SRAM, yielding a more accurate initialization that enables fast forward and backward computation. Combined with a sparsity-aware GPU implementation that skips zero blocks with low overhead, AdaSplash-2 matches or improves per-step training time relative to FlashAttention-2 when block sparsity is moderate-to-high (e.g., $>$60\%), which often occurs at long-context lengths. On downstream tasks, models trained with our efficient $α$-entmax attention match softmax baselines at short-context lengths and achieve substantial gains in long-context settings.

cs.LG↗

arXiv2026

CuraLight: Debate-Guided Data Curation for LLM-Centered Traffic Signal Control

Traffic signal control (TSC) is a core component of intelligent transportation systems (ITS), aiming to reduce congestion, emissions, and travel time. Recent approaches based on reinforcement learning (RL) and large language models (LLMs) have improved adaptivity, but still suffer from limited interpretability, insufficient interaction data, and weak generalization to heterogeneous intersections. This paper proposes CuraLight, an LLM-centered framework where an RL agent assists the fine-tuning of an LLM-based traffic signal controller. The RL agent explores traffic environments and generates high-quality interaction trajectories, which are converted into prompt-response pairs for imitation fine-tuning. A multi-LLM ensemble deliberation system further evaluates candidate signal timing actions through structured debate, providing preference-aware supervision signals for training. Experiments conducted in SUMO across heterogeneous real-world networks from Jinan, Hangzhou, and Yizhuang demonstrate that CuraLight consistently outperforms state-of-the-art baselines, reducing average travel time by 5.34 percent, average queue length by 5.14 percent, and average waiting time by 7.02 percent. The results highlight the effectiveness of combining RL-assisted exploration with deliberation-based data curation for scalable and interpretable traffic signal control.

cs.AI↗

arXiv2026

RAM: Recover Any 3D Human Motion in-the-Wild

RAM incorporates a motion-aware semantic tracker with adaptive Kalman filtering to achieve robust identity association under severe occlusions and dynamic interactions. A memory-augmented Temporal HMR module further enhances human motion reconstruction by injecting spatio-temporal priors for consistent and smooth motion estimation. Moreover, a lightweight Predictor module forecasts future poses to maintain reconstruction continuity, while a gated combiner adaptively fuses reconstructed and predicted features to ensure coherence and robustness. Experiments on in-the-wild multi-person benchmarks such as PoseTrack and 3DPW, demonstrate that RAM substantially outperforms previous state-of-the-art in both Zero-shot tracking stability and 3D accuracy, offering a generalizable paradigm for markerless 3D human motion capture in-the-wild.

cs.CV↗

arXiv2026

Crossover Equation of State Constrained by Astronomical Observations and pQCD

The hadron--quark crossover equation of state (EOS) of neutron star (NS) matter is investigated by combining relativistic mean-field (RMF) hadronic models with the Nambu--Jona-Lasinio (NJL) model for quark matter. The vector and diquark coupling constants of the NJL model are constrained using perturbative QCD (pQCD) calculations at high density through a scale-averaging likelihood approach, together with constraints from NS observations and the causality condition on the speed of sound. It is found that the diquark coupling is tightly constrained to $H \simeq 1.5G_s$, while the vector coupling is restricted to $G_v \lesssim 1.1G_s$ by the combined pQCD and astrophysical constraints. Crossover EOSs are constructed based on three hadronic RMF parameter sets, and their thermodynamic properties, sound speed behaviour, and trace anomaly are analysed. The resulting EOSs are applied to calculate NS global and dynamical properties, including mass--radius relations, tidal deformabilities, and fundamental radial oscillation frequencies. Compared with pure hadronic EOSs, the hadron--quark crossover is shown to significantly enhance the maximum NS mass, particularly for softer hadronic EOSs, while remaining consistent with observational bounds. It is further shown that the fundamental radial oscillation frequencies predicted by different EOSs exhibit pronounced differences, especially for intermediate-mass NSs, indicating that radial modes may provide a sensitive probe of the internal composition of NSs. These results indicate that quantitative NS observables may provide potential signatures of quark matter in NS interiors.

nucl-th↗

arXiv2026

Unsupervised Multi-agent and Single-agent Perception from Cooperative Views

The LiDAR-based multi-agent and single-agent perception has shown promising performance in environmental understanding for robots and automated vehicles. However, there is no existing method that simultaneously solves both multi-agent and single-agent perception in an unsupervised way. By sharing sensor data between multiple agents via communication, this paper discovers two key insights: 1) Improved point cloud density after the data sharing from cooperative views could benefit unsupervised object classification, 2) Cooperative view of multiple agents can be used as unsupervised guidance for the 3D object detection in the single view. Based on these two discovered insights, we propose an Unsupervised Multi-agent and Single-agent (UMS) perception framework that leverages multi-agent cooperation without human annotations to simultaneously solve multi-agent and single-agent perception. UMS combines a learning-based Proposal Purifying Filter to better classify the candidate proposals after multi-agent point cloud density cooperation, followed by a Progressive Proposal Stabilizing module to yield reliable pseudo labels by the easy-to-hard curriculum learning. Furthermore, we design a Cross-View Consensus Learning to use multi-agent cooperative view to guide detection in single-agent view. Experimental results on two public datasets V2V4Real and OPV2V show that our UMS method achieved significantly higher 3D detection performance than the state-of-the-art methods on both multi-agent and single-agent perception tasks in an unsupervised setting.

cs.CV↗

arXiv2026

ECHO: Event-Centric Hypergraph Operations via Multi-Agent Collaboration for Multimedia Event Extraction

Multimedia event extraction (M2E2) aims to predict triggers, ground arguments across text and images, and then assemble them into schema-consistent event records. Recent LLM-based approaches have shown strong potential for M2E2, but their intermediate event hypotheses often remain implicit, and event-argument linking is still tightly coupled with role binding. This leaves little opportunity to inspect or revise intermediate event hypotheses and makes predictions brittle to early errors. To bridge this gap, we present ECHO, a multi-agent framework that reframes M2E2 as iterative refinement over an explicit Multimedia Event Hypergraph (MEHG). Instead of relying on implicit linear generation, ECHO performs auditable atomic updates over a shared hypergraph, making intermediate event structures explicit and revisable. Furthermore, we introduce a Link-then-Bind strategy that decouples event-argument linking from role binding, reducing premature semantic commitment during structured prediction. Extensive experiments on the M2E2 benchmark show that ECHO consistently outperforms prior state-of-the-art approaches, achieving gains of 7.3 and 15.5 F1 points on event mention and argument role, respectively.

cs.CV↗

arXiv2026

R2MED: A Benchmark for Reasoning-Driven Medical Retrieval

Current medical retrieval benchmarks primarily emphasize lexical or shallow semantic similarity, overlooking the reasoning-intensive demands that are central to clinical decision-making. In practice, physicians often retrieve authoritative medical evidence to support diagnostic hypotheses. Such evidence typically aligns with an inferred diagnosis rather than the surface form of a patient's symptoms, leading to low lexical or semantic overlap between queries and relevant documents. To address this gap, we introduce R2MED, the first benchmark explicitly designed for reasoning-driven medical retrieval. It comprises 876 queries spanning three tasks: Q&A reference retrieval, clinical evidence retrieval, and clinical case retrieval. These tasks are drawn from five representative medical scenarios and twelve body systems, capturing the complexity and diversity of real-world medical information needs. We evaluate 15 widely-used retrieval systems on R2MED and find that even the best model achieves only 31.4 nDCG@10, demonstrating the benchmark's difficulty. Classical re-ranking and generation-augmented retrieval methods offer only modest improvements. Although large reasoning models improve performance via intermediate inference generation, the best results still peak at 41.4 nDCG@10. These findings underscore a substantial gap between current retrieval techniques and the reasoning demands of real clinical tasks. We release R2MED as a challenging benchmark to foster the development of next-generation medical retrieval systems with enhanced reasoning capabilities. Data and code are available at https://github.com/R2MED/R2MED

cs.IR↗

arXiv2026

MPI-Q: A Message Communication Library for Large-Scale Classical-Quantum Heterogeneous Hybrid Distributed Computing

The classical-quantum system heterogeneity (different data characteristics, execution paradigms and synchronization mechanism etc.) renders existing distributed communication mechanisms (e.g. MPI, NCCL etc.) inadequate. This bottleneck severely impairs operational synergy and programming efficiency. Thus, the performance of hybrid applications on classical-quantum heterogeneous infrastructures is directly limited. To address these challenges, this paper proposes a message-passing library tailored for large-scale classical-quantum heterogeneous distributed computing, referred to as MPI-Q. The design centers on three mechanisms. First, it defines a heterogeneous hybrid communication domain that achieves unified management of classical and quantum processes in heterogeneous hybrid systems. Second, it uses a lightweight communication path that allows classical control nodes to send device-ready waveform data directly to quantum MonitorProcesses, avoiding unnecessary relay stages. Third, it establishes a heterogeneous hybrid synchronization mechanism to tackle the problem of timing control for multi-node quantum operations. While retaining the traditional MPI programming model, MPI-Q achieves extension toward quantum subsystems. Experiments on distributed GHZ state preparation demonstrate that this model exhibits near-linear scalability, achieving a maximum speedup of 18.76 times on 24 quantum nodes. This proves that the library can effectively support large-scale heterogeneous hybrid distributed computing applications, filling the technical gap in this field.

cs.DC↗

arXiv2026

Distributed User Scheduling in Multi-Cell MIMO O-RAN with QoS Constraints

Distributed scheduling is essential for open radio access network (O-RAN) employing advanced physical-layer techniques such as multi-user MIMO (MU-MIMO), carrier aggregation (CA), and joint transmission (JT). This work investigates the multi-component-carrier (multi-CC) resource block group (RBG) scheduling in MU-MIMO O-RAN with both JT and non-JT users. We formulate a scheduling optimization problem to maximize throughput subject to user-specific quality of service (QoS) requirements while ensuring consistent allocations across cooperating O-RAN radio units (O-RUs) required by JT transmission. The strong variable coupling, non-convexity, and combinatorial complexity make the problem highly challenging. To tackle this, we extend the eigen-based zero-forcing transceiver design to JT users and leverage massive MIMO asymptotic properties to derive a tractable, separable rate approximation. Building on this, we develop two solutions: a centralized block coordinate descent benchmark and a distributed scheduler aligned with the O-RAN architecture. The proposed distributed scheme achieves near-centralized performance with only one round of lightweight coordination among cells, significantly reducing complexity and delay. Extensive simulations validate that our distributed scheduler achieves high scalability, fast convergence, and better QoS satisfaction rate in large-scale MU-MIMO networks.

eess.SP↗

arXiv2026

Toward Distributed User Scheduling and Coordinated Beamforming in Multi-Cell mmWave Networks: A Sensing-Assisted Framework

Providing guaranteed quality of service for cell-edge users remains a longstanding challenge in wireless networks. While coordinated interference management was proposed decades ago, its potential has been limited by computational complexity and backhaul resource constraints. Distributed user scheduling and coordinated beamforming (D-USCB) offers a scalable solution but faces practical challenges in acquiring inter-cell channel state information (CSI), as base stations (BSs) are often restricted to signal strength measurements, and high-dimensional CSI exchange incurs substantial overhead. Inspired by integrated sensing and communication (ISAC), this paper proposes a sensing-assisted D-USCB (SD-USCB) framework to maximize the network throughput of multi-cell mmWave networks. Firstly, the framework leverages channel knowledge maps (CKMs) that map user locations to CSI estimates, where user locations are proactively sensed via ISAC echoes. Secondly, we employ a signal-to-average-leakage-plus-interference-plus-noise ratio (SALINR) metric for distributed ISAC beamforming optimization, in which BSs simultaneously communicate with users and sense their locations. These two components jointly enable distributed coordinated transmission with only user location information exchanged among BSs, thereby substantially reducing backhaul overhead. In addition, we devise efficient distributed user scheduling and ISAC beamforming algorithms to jointly optimize communication and sensing performance. Extensive numerical results demonstrate significant improvements in network throughput, validating the efficacy of the proposed framework.

eess.SP↗

arXiv2026

Seen2Scene: Completing Realistic 3D Scenes with Visibility-Guided Flow

We present Seen2Scene, the first flow matching-based approach that trains directly on incomplete, real-world 3D scans for scene completion and generation. Unlike prior methods that rely on complete and hence synthetic 3D data, our approach introduces visibility-guided flow matching, which explicitly masks out unknown regions in real scans, enabling effective learning from real-world, partial observations. We represent 3D scenes using truncated signed distance field (TSDF) volumes encoded in sparse grids and employ a sparse transformer to efficiently model complex scene structures while masking unknown regions. We employ 3D layout boxes as an input conditioning signal, and our approach is flexibly adapted to various other inputs such as text or partial scans. By learning directly from real-world, incomplete 3D scans, Seen2Scene enables realistic 3D scene completion for complex, cluttered real environments. Experiments demonstrate that our model produces coherent, complete, and realistic 3D scenes, outperforming baselines in completion accuracy and generation quality.

cs.CV↗

arXiv2026

Quasi-random splitting method for accurate and efficient multiphysics simulation

We propose a quasi-random operator splitting method for evolution equations driven by multiple mechanisms. The method uses a low-discrepancy sequence to generate the ordering of the subflows, while requiring only one application of each subflow per time step. In particular, for a decomposition into $p$ operators, the classical multi-operator Strang splitting requires essentially $2p-2$ subflow evaluations per step, whereas the present method uses only $p$. In contrast to randomized splitting, the quasi-random scheme is deterministic once the underlying sequence is fixed, so its improved accuracy is achieved in a single run rather than through averaging over many independent realizations. To analyze this method, we develop a convergence framework that exploits the discrepancy structure of the induced ordering sequence and translates it into cancellation in the accumulated local errors. For two operators, this yields an essentially second-order global error bound of order $O(τ^{2}|\log τ|)$ for bounded linear problems. We further extend the analysis to the Allen--Cahn equation and present numerical experiments, including bounded linear systems and the Allen--Cahn equation, which confirm the predicted convergence behavior and demonstrate that the proposed method achieves near-Strang accuracy at a substantially lower computational cost.

math.NA↗