arXiv subjects

Weikang Li

Publications and source records attributed to Weikang Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

CMRAG: Co-modality-based visual document retrieval and question answering

Retrieval-Augmented Generation (RAG) has become a core paradigm in document question answering tasks. However, existing methods have limitations when dealing with multimodal documents: one category of methods relies on layout analysis and text extraction, which can only utilize explicit text information and struggle to capture images or unstructured content; the other category treats document segmentation as visual input and directly passes it to visual language models (VLMs) for processing, yet it ignores the semantic advantages of text, leading to suboptimal retrieval and generation results. To address these research gaps, we propose the Co-Modality-based RAG (CMRAG) framework, which can simultaneously leverage texts and images for more accurate retrieval and generation. Our framework includes two key components: (1) a Unified Encoding Model (UEM) that projects queries, parsed text, and images into a shared embedding space via triplet-based training, and (2) a Unified Co-Modality-informed Retrieval (UCMR) method that statistically normalizes similarity scores to effectively fuse cross-modal signals. To support research in this direction, we further construct and release a large-scale triplet dataset of (query, text, image) examples. Experiments demonstrate that our proposed framework consistently outperforms single-modality--based RAG in multiple visual document question-answering (VDQA) benchmarks. The findings of this paper show that integrating co-modality information into the RAG framework in a unified manner is an effective approach to improving the performance of complex VDQA systems.

cs.CL↗

arXiv2026

DuCCAE: A Hybrid Engine for Immersive Conversation via Collaboration, Augmentation, and Evolution

Immersive conversational systems in production face a persistent trade-off between responsiveness and long-horizon task capability. Real-time interaction is achievable for lightweight turns, but requests involving planning and tool invocation (e.g., search and media generation) produce heavy-tail execution latency that degrades turn-taking, persona consistency, and user trust. To address this challenge, we propose DuCCAE (Conversation while Collaboration with Augmentation and Evolution), a hybrid engine for immersive conversation deployed within Baidu Search, serving millions of users. DuCCAE decouples real-time response generation from asynchronous agentic execution and synchronizes them via a shared state that maintains session context and execution traces, enabling asynchronous results to be integrated back into the ongoing dialogue. The system orchestrates five subsystems-Info, Conversation, Collaboration, Augmentation, and Evolution-to support multi-agent collaboration and continuous improvement. We evaluate DuCCAE through a comprehensive framework that combines offline benchmarking on the Du-Interact dataset and large-scale production evaluation within Baidu Search. Experimental results demonstrate that DuCCAE outperforms strong baselines in agentic execution reliability and dialogue quality while reducing latency to fit strict real-time budgets. Crucially, deployment metrics since June 2025 confirm substantial real-world effectiveness, evidenced by a tripling of Day-7 user retention to 34.2% and a surge in the complex task completion rate to 65.2%. Our hybrid architecture successfully preserves conversational continuity while enabling reliable agentic execution, offering practical guidelines for deploying scalable agentic systems in industrial settings.

cs.CL↗

arXiv2026

LM-Lexicon: Improving Definition Modeling via Harmonizing Semantic Experts

We introduce LM-Lexicon, an innovative definition modeling approach that incorporates data clustering, semantic expert learning, and model merging using a sparse mixture-of-experts architecture. By decomposing the definition modeling task into specialized semantic domains, where small language models are trained as domain experts, LM-Lexicon achieves substantial improvements (+7% BLEU score compared with the prior state-of-the-art model) over existing methods on five widely used benchmarks. Empirically, we demonstrate that 1) the clustering strategy enables fine-grained expert specialization with nearly 10% improvement in definition quality; 2) the semantic-aware domain-level routing mechanism achieves higher expert efficacy (+1%) than conventional token-level routing; and 3) further performance gains can be obtained through test-time compute and semantic expert scaling. Our work advances definition modeling while providing insights into the development of efficient language models for semantic-intensive applications.

cs.CL↗

arXiv2026

Beyond Alignment: Expanding Reasoning Capacity via Manifold-Reshaping Policy Optimization

Reinforcement Learning with Verifiable Rewards (RLVR) has demonstrated remarkable success in enhancing the reasoning capabilities of Large Language Models (LLMs). However, recent studies question whether RL genuinely expands reasoning capacity or merely aligns existing latent capabilities, arguing that exploration remains confined within the pre-trained model's low-rank bias manifold. In this work, we challenge this accessibility boundary hypothesis by demonstrating that the latent reasoning space can be fundamentally expanded through targeted geometric interventions. We propose Manifold-Reshaping Policy Optimization (MRPO), a geometric framework designed to fundamentally restructure the inference space of LLMs. MRPO operates in two stages: first, we employ Spectral Orthogonal Exploration (SOE) to eject the policy initialization into the null space of the bias manifold; second, we integrate an Effective Rank regularization term into the policy optimization objective. This approach incentivizes the discovery and maintenance of high-dimensional reasoning trajectories against the entropy-reducing tendency of standard RL. Empirically, our 4B-parameter method achieves state-of-the-art performance on mathematical tasks, significantly outperforming larger models (e.g., Qwen3-32B) and expanding the capability boundary beyond standard GRPO. Our code is available at https://anonymous.4open.science/r/MRPO-D57B/

cs.LG↗

arXiv2026

Stay in Character, Stay Safe: Dual-Cycle Adversarial Self-Evolution for Safety Role-Playing Agents

LLM-based role-playing has rapidly improved in fidelity, yet stronger adherence to persona constraints commonly increases vulnerability to jailbreak attacks, especially for risky or negative personas. Most prior work mitigates this issue with training-time solutions (e.g., data curation or alignment-oriented regularization). However, these approaches are costly to maintain as personas and attack strategies evolve, can degrade in-character behavior, and are typically infeasible for frontier closed-weight LLMs. We propose a training-free Dual-Cycle Adversarial Self-Evolution framework with two coupled cycles. A Persona-Targeted Attacker Cycle synthesizes progressively stronger jailbreak prompts, while a Role-Playing Defender Cycle distills observed failures into a hierarchical knowledge base of (i) global safety rules, (ii) persona-grounded constraints, and (iii) safe in-character exemplars. At inference time, the Defender retrieves and composes structured knowledge from this hierarchy to guide generation, producing responses that remain faithful to the target persona while satisfying safety constraints. Extensive experiments across multiple proprietary LLMs show consistent gains over strong baselines on both role fidelity and jailbreak resistance, and robust generalization to unseen personas and attack prompts.

cs.AI↗

arXiv2026

Video-MSR: Benchmarking Multi-hop Spatial Reasoning Capabilities of MLLMs

Spatial reasoning has emerged as a critical capability for Multimodal Large Language Models (MLLMs), drawing increasing attention and rapid advancement. However, existing benchmarks primarily focus on single-step perception-to-judgment tasks, leaving scenarios requiring complex visual-spatial logical chains significantly underexplored. To bridge this gap, we introduce Video-MSR, the first benchmark specifically designed to evaluate Multi-hop Spatial Reasoning (MSR) in dynamic video scenarios. Video-MSR systematically probes MSR capabilities through four distinct tasks: Constrained Localization, Chain-based Reference Retrieval, Route Planning, and Counterfactual Physical Deduction. Our benchmark comprises 3,052 high-quality video instances with 4,993 question-answer pairs, constructed via a scalable, visually-grounded pipeline combining advanced model generation with rigorous human verification. Through a comprehensive evaluation of 20 state-of-the-art MLLMs, we uncover significant limitations, revealing that while models demonstrate proficiency in surface-level perception, they exhibit distinct performance drops in MSR tasks, frequently suffering from spatial disorientation and hallucination during multi-step deductions. To mitigate these shortcomings and empower models with stronger MSR capabilities, we further curate MSR-9K, a specialized instruction-tuning dataset, and fine-tune Qwen-VL, achieving a +7.82% absolute improvement on Video-MSR. Our results underscore the efficacy of multi-hop spatial instruction data and establish Video-MSR as a vital foundation for future research. The code and data will be available at https://github.com/ruiz-nju/Video-MSR.

cs.CV↗

arXiv2026

Decide Then Retrieve: A Training-Free Framework with Uncertainty-Guided Triggering and Dual-Path Retrieval

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating external knowledge, but existing approaches indiscriminately trigger retrieval and rely on single-path evidence construction, often introducing noise and limiting performance gains. In this work, we propose Decide Then Retrieve (DTR), a training-free framework that adaptively determines when retrieval is necessary and how external information should be selected. DTR leverages generation uncertainty to guide retrieval triggering and introduces a dual-path retrieval mechanism with adaptive information selection to better handle sparse and ambiguous queries. Extensive experiments across five open-domain QA benchmarks, multiple model scales, and different retrievers demonstrate that DTR consistently improves EM and F1 over standard RAG and strong retrieval-enhanced baselines, while reducing unnecessary retrievals. The code and data used in this paper are available at https://github.com/ChenWangHKU/DTR.

cs.CL↗

arXiv2026

A Unified Frequency Principle for Quantum and Classical Machine Learning

Quantum neural networks constitute a key class of near-term quantum learning models, yet their training dynamics remain not fully understood. Here, we present a unified theoretical framework for the frequency principle (F-principle) that characterizes the training dynamics of both classical and quantum neural networks. Within this framework, we prove that quantum neural networks exhibit a spectral bias toward learning low-frequency components of target functions, mirroring the behavior observed in classical deep networks. We further analyze the impact of noise and show that, when single-qubit noise is applied after encoding-layer rotations and modeled as a Pauli channel aligned with the rotation axis, the Fourier component labeled by $\boldsymbolω$ is suppressed by a factor $(1-2γ)^{\|\boldsymbolω\|_1}$. This leads to exponential attenuation of high-frequency terms while preserving the learnability of low-frequency structure. In the same setting, we establish that the resulting noisy circuits admit efficient classical simulation up to average-case error. Numerical experiments corroborate our theoretical predictions: Quantum neural networks primarily learn low-frequency features during early optimization and maintain robustness against dephasing and depolarizing noise acting on the encoding layer. Our results provide a frequency-domain lens that unifies classical and quantum learning dynamics, clarifies the role of noise in shaping trainability, and guides the design of noise-resilient quantum neural networks.

quant-ph↗

arXiv2025

FingerCap: Fine-grained Finger-level Hand Motion Captioning

Understanding fine-grained human hand motion is fundamental to visual perception, embodied intelligence, and multimodal communication. In this work, we propose Fine-grained Finger-level Hand Motion Captioning (FingerCap), which aims to generate textual descriptions that capture detailed finger-level semantics of hand actions. To support this task, we curate FingerCap-40K, a large-scale corpus of 40K paired hand-motion videos and captions spanning two complementary sources: concise instruction-style finger motions and diverse, naturalistic hand-object interactions. To enable effective evaluation, we employ HandJudge, a LLM-based rubric that measures finger-level correctness and motion completeness. Temporal sparsity remains a fundamental bottleneck for current Video-MLLMs, since sparse RGB sampling is insufficient to capture the subtle, high-frequency dynamics underlying fine finger motions. As a simple and compute-friendly remedy, we introduce FiGOP (Finger Group-of-Pictures), which pairs each RGB keyframe with subsequent hand keypoints until the next keyframe. A lightweight temporal encoder converts the keypoints into motion embeddings and integrates them with RGB features. FiGOP adapts the classic GOP concept to finger motion, recovering fine temporal cues without increasing RGB density. Experiments on FingerCap-40K show that strong open- and closed-source Video-MLLMs still struggle with finger-level reasoning, while our FiGOP-augmented model yield consistent gains under HandJudge and human studies.

cs.CV↗

arXiv2025

Probabilistic Modeling of Intentions in Socially Intelligent LLM Agents

We present a probabilistic intent modeling framework for large language model (LLM) agents in multi-turn social dialogue. The framework maintains a belief distribution over a partner's latent intentions, initialized from contextual priors and dynamically updated through likelihood estimation after each utterance. The evolving distribution provides additional contextual grounding for the policy, enabling adaptive dialogue strategies under uncertainty. Preliminary experiments in the SOTOPIA environment show consistent improvements: the proposed framework increases the Overall score by 9.0% on SOTOPIA-All and 4.1% on SOTOPIA-Hard compared with the Qwen2.5-7B baseline, and slightly surpasses an oracle agent that directly observes partner intentions. These early results suggest that probabilistic intent modeling can contribute to the development of socially intelligent LLM agents.

cs.AI↗

arXiv2025

Cross-LoRA: A Data-Free LoRA Transfer Framework across Heterogeneous LLMs

Traditional parameter-efficient fine-tuning (PEFT) methods such as LoRA are tightly coupled with the base model architecture, which constrains their applicability across heterogeneous pretrained large language models (LLMs). To address this limitation, we introduce Cross-LoRA, a data-free framework for transferring LoRA modules between diverse base models without requiring additional training data. Cross-LoRA consists of two key components: (a) LoRA-Align, which performs subspace alignment between source and target base models through rank-truncated singular value decomposition (SVD) and Frobenius-optimal linear transformation, ensuring compatibility under dimension mismatch; and (b) LoRA-Shift, which applies the aligned subspaces to project source LoRA weight updates into the target model parameter space. Both components are data-free, training-free, and enable lightweight adaptation on a commodity GPU in 20 minutes. Experiments on ARCs, OBOA and HellaSwag show that Cross-LoRA achieves relative gains of up to 5.26% over base models. Across other commonsense reasoning benchmarks, Cross-LoRA maintains performance comparable to that of directly trained LoRA adapters.

cs.LG↗

arXiv2025

PAIRS: Parametric-Verified Adaptive Information Retrieval and Selection for Efficient RAG

Retrieval-Augmented Generation (RAG) has become a cornerstone technique for enhancing large language models (LLMs) with external knowledge. However, current RAG systems face two critical limitations: (1) they inefficiently retrieve information for every query, including simple questions that could be resolved using the LLM's parametric knowledge alone, and (2) they risk retrieving irrelevant documents when queries contain sparse information signals. To address these gaps, we introduce Parametric-verified Adaptive Information Retrieval and Selection (PAIRS), a training-free framework that integrates parametric and retrieved knowledge to adaptively determine whether to retrieve and how to select external information. Specifically, PAIRS employs a dual-path generation mechanism: First, the LLM produces both a direct answer and a context-augmented answer using self-generated pseudo-context. When these outputs converge, PAIRS bypasses external retrieval entirely, dramatically improving the RAG system's efficiency. For divergent cases, PAIRS activates a dual-path retrieval (DPR) process guided by both the original query and self-generated contextual signals, followed by an Adaptive Information Selection (AIS) module that filters documents through weighted similarity to both sources. This simple yet effective approach can not only enhance efficiency by eliminating unnecessary retrievals but also improve accuracy through contextually guided retrieval and adaptive information selection. Experimental results on six question-answering (QA) benchmarks show that PAIRS reduces retrieval costs by around 25% (triggering for only 75% of queries) while still improving accuracy-achieving +1.1% EM and +1.0% F1 over prior baselines on average.

cs.CL↗

arXiv2025

Solving excited states for long-range interacting trapped ions with neural networks

The computation of excited states in strongly interacting quantum many-body systems is of fundamental importance. Yet, it is notoriously challenging due to the exponential scaling of the Hilbert space dimension with the system size. Here, we introduce a neural network-based algorithm that can simultaneously output multiple low-lying excited states of a quantum many-body spin system in an accurate and efficient fashion. This algorithm, dubbed the neural quantum excited-state (NQES) algorithm, requires no explicit orthogonalization of the states and is generally applicable to higher dimensions. We demonstrate, through concrete examples including the Haldane-Shastry model with all-to-all interactions, that the NQES algorithm is capable of efficiently computing multiple excited states and their related observable expectations. In addition, we apply the NQES algorithm to two classes of long-range interacting trapped-ion systems in a two-dimensional Wigner crystal. For non-decaying all-to-all interactions with alternating signs, our computed low-lying excited states bear spatial correlation patterns similar to those of the ground states, which closely match recent experimental observations that the quasi-adiabatically prepared state accurately reproduces analytical ground-state correlations. For a system of up to 300 ions with power-law decaying antiferromagnetic interactions, we successfully uncover its gap scaling and correlation features. Our results establish a scalable and efficient algorithm for computing excited states of interacting quantum many-body systems, which holds potential applications ranging from benchmarking quantum devices to photoisomerization.

quant-ph↗

arXiv2025

Demonstration of low-overhead quantum error correction codes

Quantum computers hold the potential to surpass classical computers in solving complex computational problems. However, the fragility of quantum information and the error-prone nature of quantum operations make building large-scale, fault-tolerant quantum computers a prominent challenge. To combat errors, pioneering experiments have demonstrated a variety of quantum error correction codes. Yet, most of these codes suffer from low encoding efficiency, and their scalability is hindered by prohibitively high resource overheads. Here, we report the demonstration of two low-overhead quantum low-density parity-check (qLDPC) codes, a distance-4 bivariate bicycle code and a distance-3 qLDPC code, on our latest superconducting processor, Kunlun, featuring 32 long-range-coupled transmon qubits. Utilizing a two-dimensional architecture with overlapping long-range couplers, we demonstrate simultaneous measurements of all nonlocal weight-6 stabilizers via the periodic execution of an efficient syndrome extraction circuit. We achieve a logical error rate per logical qubit per cycle of $(8.91 \pm 0.17)\%$ for the distance-4 bivariate bicycle code with four logical qubits and $(7.77 \pm 0.12)\%$ for the distance-3 qLDPC code with six logical qubits. Our results establish the feasibility of implementing various qLDPC codes with long-range coupled superconducting processors, marking a crucial step towards large-scale low-overhead quantum error correction.

quant-ph↗

arXiv2025

Quantum automated learning with provable and explainable trainability

Machine learning is widely believed to be one of the most promising practical applications of quantum computing. Existing quantum machine learning schemes typically employ a quantum-classical hybrid approach that relies crucially on gradients of model parameters. Such an approach lacks provable convergence to global minima and will become infeasible as quantum learning models scale up. Here, we introduce quantum automated learning, where no variational parameter is involved and the training process is converted to quantum state preparation. In particular, we encode training data into unitary operations and iteratively evolve a random initial state under these unitaries and their inverses, with a target-oriented perturbation towards higher prediction accuracy sandwiched in between. Under reasonable assumptions, we rigorously prove that the evolution converges exponentially to the desired state corresponding to the global minimum of the loss function. We show that such a training process can be understood from the perspective of preparing quantum states by imaginary time evolution, where the data-encoded unitaries together with target-oriented perturbations would train the quantum learning model in an automated fashion. We further prove that the quantum automated learning paradigm features good generalization ability with the generalization error upper bounded by the ratio between a logarithmic function of the Hilbert space dimension and the number of training samples. In addition, we carry out extensive numerical simulations on real-life images and quantum data to demonstrate the effectiveness of our approach and validate the assumptions. Our results establish an unconventional quantum learning strategy that is gradient-free with provable and explainable trainability, which would be crucial for large-scale practical applications of quantum computing in machine learning scenarios.

quant-ph↗

arXiv2025

Observation of topological prethermal strong zero modes

Symmetry-protected topological phases cannot be described by any local order parameter and are beyond the conventional symmetry-breaking paradigm for understanding quantum matter. They are characterized by topological boundary states robust against perturbations that respect the protecting symmetry. In a clean system without disorder, these edge modes typically only occur for the ground states of systems with a bulk energy gap and would not survive at finite temperatures due to mobile thermal excitations. Here, we report the observation of a distinct type of topological edge modes, which are protected by emergent symmetries and persist even up to infinite temperature, with an array of 100 programmable superconducting qubits. In particular, through digital quantum simulation of the dynamics of a one-dimensional disorder-free "cluster" Hamiltonian, we observe robust long-lived topological edge modes over up to 30 cycles at a wide range of temperatures. By monitoring the propagation of thermal excitations, we show that despite the free mobility of these excitations, their interactions with the edge modes are substantially suppressed in the dimerized regime due to an emergent U(1)$\times$U(1) symmetry, resulting in an unusually prolonged lifetime of the topological edge modes even at infinite temperature. In addition, we exploit these topological edge modes as logical qubits and prepare a logical Bell state, which exhibits persistent coherence in the dimerized and off-resonant regime, despite the system being disorder-free and far from its ground state. Our results establish a viable digital simulation approach to experimentally exploring a variety of finite-temperature topological phases and demonstrate a potential route to construct long-lived robust boundary qubits that survive to infinite temperature in disorder-free systems.

quant-ph↗

arXiv2024

Quantum delegated and federated learning via quantum homomorphic encryption

Quantum learning models hold the potential to bring computational advantages over the classical realm. As powerful quantum servers become available on the cloud, ensuring the protection of clients' private data becomes crucial. By incorporating quantum homomorphic encryption schemes, we present a general framework that enables quantum delegated and federated learning with a computation-theoretical data privacy guarantee. We show that learning and inference under this framework feature substantially lower communication complexity compared with schemes based on blind quantum computing. In addition, in the proposed quantum federated learning scenario, there is less computational burden on local quantum devices from the client side, since the server can operate on encrypted quantum data without extracting any information. We further prove that certain quantum speedups in supervised learning carry over to private delegated learning scenarios employing quantum kernel methods. Our results provide a valuable guide toward privacy-guaranteed quantum learning on the cloud, which may benefit future studies and security-related applications.

quant-ph↗

arXiv2024

Quantum continual learning on a programmable superconducting processor

Quantum computers may outperform classical computers on machine learning tasks. In recent years, a variety of quantum algorithms promising unparalleled potential to enhance, speed up, or innovate machine learning have been proposed. Yet, quantum learning systems, similar to their classical counterparts, may likewise suffer from the catastrophic forgetting problem, where training a model with new tasks would result in a dramatic performance drop for the previously learned ones. This problem is widely believed to be a crucial obstacle to achieving continual learning of multiple sequential tasks. Here, we report an experimental demonstration of quantum continual learning on a fully programmable superconducting processor. In particular, we sequentially train a quantum classifier with three tasks, two about identifying real-life images and the other on classifying quantum states, and demonstrate its catastrophic forgetting through experimentally observed rapid performance drops for prior tasks. To overcome this dilemma, we exploit the elastic weight consolidation strategy and show that the quantum classifier can incrementally learn and retain knowledge across the three distinct tasks, with an average prediction accuracy exceeding 92.3%. In addition, for sequential tasks involving quantum-engineered data, we demonstrate that the quantum classifier can achieve a better continual learning performance than a commonly used classical feedforward network with a comparable number of variational parameters. Our results establish a viable strategy for empowering quantum learning systems with desirable adaptability to multiple sequential tasks, marking an important primary experimental step towards the long-term goal of achieving quantum artificial general intelligence.

quant-ph↗