arXiv subjects

Qianlong Wang

Publications and source records attributed to Qianlong Wang.

14 recordsLinked to original sources

arXiv2026

Janus: An Algorithm-Evaluator Co-Evolution Framework for LLM-Driven Discovery under Expensive Evaluation Budgets

LLM-driven program discovery relies on rapid evaluator feedback, but many scientific and engineering tasks require high-fidelity simulations, hardware execution, or physical experiments, making each evaluation expensive. Cheap surrogate evaluators can reduce this cost, yet fixed surrogates are vulnerable to search-induced distribution shift and are difficult to fit reliably from sparse, search-biased labels. We introduce Janus, a framework that uses LLMs to co-evolve target programs and executable proxy evaluators. To address label scarcity, Janus leverages domain knowledge encoded in LLMs to generate task-specific evaluator programs and calibrates them using real outcomes. To mitigate distribution shift, Janus evolves evaluators alongside target programs, selects them using a promotion-aligned objective, and maintains region-conditioned portfolios with online credit updates. Because proxy predictions remain fallible, Janus uses them only to prioritize candidates and requires real validation before candidates can enter the target-program population or update the incumbent. Across five scientific and engineering design tasks, Janus achieves a larger area under the best-so-far improvement curve over the real-evaluation budget and higher final performance than a matched baseline that evolves only target programs. On average, Janus reaches 99/% of the baseline's final improvement with 59.1/% fewer real evaluations. Evolved proxy evaluators also rank promising candidates more accurately than their seed versions. Together, these results extend evaluator-guided LLM discovery from tasks with cheap, scalable feedback to scientific domains where trustworthy evaluation is scarce and expensive.

cs.AI↗

arXiv2026

Strong coupling of virtual negative states in the Kapitza-Dirac effect

Negative states are an intrinsic property of relativistic quantum theory and related to anti-particles in the context of the Dirac sea concept. We show that negative states can dominantly contribute to the diffraction amplitude in the quantum dynamics of the two-photon Kapitza-Dirac effect. We draw our conclusion by investigating solutions from time-dependent perturbation theory, where the perturbative solutions are in match with numeric solutions of the relativistic quantum system and also with the numeric and analytic solutions from the relativistic equations of motion of a classical point-like electron in an external standing wave light field. While our numeric solutions assume a strong laser field, the analytic solutions indicate that negative state coupling remains dominant for arbitrary low field amplitudes, where in the single-photon case (Compton scattering) negative state coupling can be mathematically associated with the interaction of a virtual electron-positron pair in the context of a quantized theory in old-fashioned perturbation theory.

quant-ph↗

arXiv2026

Rethinking Fusion: Disentangled Learning of Shared and Modality-Specific Information for Stance Detection

Multi-modal stance detection (MSD) aims to determine an author's stance toward a given target using both textual and visual content. While recent methods leverage multi-modal fusion and prompt-based learning, most fail to distinguish between modality-specific signals and cross-modal evidence, leading to suboptimal performance. We propose DiME (Disentangled Multi-modal Experts), a novel architecture that explicitly separates stance information into textual-dominant, visual-dominant, and cross-modal shared components. DiME first uses a target-aware Chain-of-Thought prompt to generate reasoning-guided textual input. Then, dual encoders extract modality features, which are processed by three expert modules with specialized loss functions: contrastive learning for modality-specific experts and cosine alignment for shared representation learning. A gating network adaptively fuses expert outputs for final prediction. Experiments on four benchmark datasets show that DiME consistently outperforms strong unimodal and multi-modal baselines under both in-target and zero-shot settings.

cs.MM↗

arXiv2026

Retrieve-Refine-Calibrate: A Framework for Complex Claim Fact-Checking

Fact-checking aims to verify the truthfulness of a claim based on the retrieved evidence. Existing methods typically follow a decomposition paradigm, in which a claim is broken down into sub-claims that are individually verified. However, the decomposition paradigm may introduce noise to the verification process due to irrelevant entities or evidence, ultimately degrading verification accuracy. To address this problem, we propose a Retrieve-Refine-Calibrate (RRC) framework based on large language models (LLMs). Specifically, the framework first identifies the entities mentioned in the claim and retrieves evidence relevant to them. Then, it refines the retrieved evidence based on the claim to reduce irrelevant information. Finally, it calibrates the verification process by re-evaluating low-confidence predictions. Experiments on two popular fact-checking datasets (HOVER and FEVEROUS-S) demonstrate that our framework achieves superior performance compared with competitive baselines.

cs.CL↗

arXiv2025

CoreEval: Automatically Building Contamination-Resilient Datasets with Real-World Knowledge toward Reliable LLM Evaluation

Data contamination poses a significant challenge to the fairness of LLM evaluations in natural language processing tasks by inadvertently exposing models to test data during training. Current studies attempt to mitigate this issue by modifying existing datasets or generating new ones from freshly collected information. However, these methods fall short of ensuring contamination-resilient evaluation, as they fail to fully eliminate pre-existing knowledge from models or preserve the semantic complexity of the original datasets. To address these limitations, we propose \textbf{CoreEval}, a \textbf{Co}ntamination-\textbf{re}silient \textbf{Eval}uation strategy for automatically updating data with real-world knowledge. This approach begins by extracting entity relationships from the original data and leveraging the GDELT database to retrieve relevant, up-to-date knowledge. The retrieved knowledge is then recontextualized and integrated with the original data, which is refined and restructured to ensure semantic coherence and enhanced task relevance. Ultimately, a robust data reflection mechanism is employed to iteratively verify and refine labels, ensuring consistency between the updated and original datasets. Extensive experiments on updated datasets validate the robustness of CoreEval, demonstrating its effectiveness in mitigating performance overestimation caused by data contamination.

cs.CL↗

arXiv2025

Targeted Distillation for Sentiment Analysis

This paper explores targeted distillation methods for sentiment analysis, aiming to build compact and practical models that preserve strong and generalizable sentiment analysis capabilities. To this end, we conceptually decouple the distillation target into knowledge and alignment and accordingly propose a two-stage distillation framework. Moreover, we introduce SentiBench, a comprehensive and systematic sentiment analysis benchmark that covers a diverse set of tasks across 12 datasets. We evaluate a wide range of models on this benchmark. Experimental results show that our approach substantially enhances the performance of compact models across diverse sentiment analysis tasks, and the resulting models demonstrate strong generalization to unseen tasks, showcasing robust competitiveness against existing small-scale models.

cs.CL↗

arXiv2025

Comprehensive and Efficient Distillation for Lightweight Sentiment Analysis Models

Recent efforts leverage knowledge distillation techniques to develop lightweight and practical sentiment analysis models. These methods are grounded in human-written instructions and large-scale user texts. Despite the promising results, two key challenges remain: (1) manually written instructions are limited in diversity and quantity, making them insufficient to ensure comprehensive coverage of distilled knowledge; (2) large-scale user texts incur high computational cost, hindering the practicality of these methods. To this end, we introduce CompEffDist, a comprehensive and efficient distillation framework for sentiment analysis. Our framework consists of two key modules: attribute-based automatic instruction construction and difficulty-based data filtering, which correspondingly tackle the aforementioned challenges. Applying our method across multiple model series (Llama-3, Qwen-3, and Gemma-3), we enable 3B student models to match the performance of 20x larger teacher models on most tasks. In addition, our approach greatly outperforms baseline methods in data efficiency, attaining the same performance level with only 10% of the data.

cs.CL↗

arXiv2024

Distilling Fine-grained Sentiment Understanding from Large Language Models

Fine-grained sentiment analysis (FSA) aims to extract and summarize user opinions from vast opinionated text. Recent studies demonstrate that large language models (LLMs) possess exceptional sentiment understanding capabilities. However, directly deploying LLMs for FSA applications incurs high inference costs. Therefore, this paper investigates the distillation of fine-grained sentiment understanding from LLMs into small language models (SLMs). We prompt LLMs to examine and interpret the sentiments of given reviews and then utilize the generated content to pretrain SLMs. Additionally, we develop a comprehensive FSA benchmark to evaluate both SLMs and LLMs. Extensive experiments on this benchmark reveal that: (1) distillation significantly enhances the performance of SLMs in FSA tasks, achieving a 6.00\% improvement in $F_1$-score, and the distilled model can outperform Llama-2-7b with only 220M parameters; (2) distillation equips SLMs with excellent zero-shot sentiment classification capabilities, enabling them to match or even exceed their teacher models. These results suggest that distillation from LLMs is a highly promising direction for FSA. We will release our code, data, and pretrained model weights at https://github.com/HITSZ-HLT/FSA-Distillation.

cs.CL↗

arXiv2024

DS$^2$-ABSA: Dual-Stream Data Synthesis with Label Refinement for Few-Shot Aspect-Based Sentiment Analysis

Recently developed large language models (LLMs) have presented promising new avenues to address data scarcity in low-resource scenarios. In few-shot aspect-based sentiment analysis (ABSA), previous efforts have explored data augmentation techniques, which prompt LLMs to generate new samples by modifying existing ones. However, these methods fail to produce adequately diverse data, impairing their effectiveness. Besides, some studies apply in-context learning for ABSA by using specific instructions and a few selected examples as prompts. Though promising, LLMs often yield labels that deviate from task requirements. To overcome these limitations, we propose DS$^2$-ABSA, a dual-stream data synthesis framework targeted for few-shot ABSA. It leverages LLMs to synthesize data from two complementary perspectives: \textit{key-point-driven} and \textit{instance-driven}, which effectively generate diverse and high-quality ABSA samples in low-resource settings. Furthermore, a \textit{label refinement} module is integrated to improve the synthetic labels. Extensive experiments demonstrate that DS$^2$-ABSA significantly outperforms previous few-shot ABSA solutions and other LLM-oriented data generation methods.

cs.CL↗

arXiv2024

Improving In-Context Learning with Prediction Feedback for Sentiment Analysis

Large language models (LLMs) have achieved promising results in sentiment analysis through the in-context learning (ICL) paradigm. However, their ability to distinguish subtle sentiments still remains a challenge. Inspired by the human ability to adjust understanding via feedback, this paper enhances ICL by incorporating prior predictions and feedback, aiming to rectify sentiment misinterpretation of LLMs. Specifically, the proposed framework consists of three steps: (1) acquiring prior predictions of LLMs, (2) devising predictive feedback based on correctness, and (3) leveraging a feedback-driven prompt to refine sentiment understanding. Experimental results across nine sentiment analysis datasets demonstrate the superiority of our framework over conventional ICL methods, with an average F1 improvement of 5.95%.

cs.CL↗

arXiv2023

Linear Absorption Spectrum of the Spin-Boson Model Studied by Extended Hierarchical Equations of Motion

With a decomposition scheme for the bath correlation function, the hierarchical equations of motion (HEOM) are extended to the zero-temperature sub-Ohmic and Ohmic spin-boson model. We investigate the linear absorption spectrum of the sub-Ohmic and Ohmic spin-boson model at zero temperature. By applying the extended HEOM, the equilibrium spin dynamics are obtained approximately. Then the linear response function is calculated according to the Kubo formula rewritten in Liouville space. To explore the essence of phase transition, we compute the linear absorption spectrum defined by the linear response function of a dipole moment. By analyzing the peak structure of the linear response spectrum, we get the dependence of linear absorption spectrum with Kondo parameter and different bath exponents. The spin relaxation dynamics are also calculated to explore the coherent-incoherent dynamic transition (CI) and delocalizedlocalized phase transition (DL). The corresponding phase diagram of DL and CI transition are also obtained. We pose a energy level picture to understand the different mechanism of DL phase transition between the deep sub-Ohmic and Ohmic spin-boson model.

quant-ph↗

arXiv2022

Energy-Efficient Computation Offloading in MobileEdge Computing Systems with Uncertainties

Computation offloading is indispensable for mobile edge computing (MEC). It uses edge resources to enable intensive computations and save energy for resource-constrained devices. Existing works generally impose strong assumptions on radio channels and network queue sizes. However, practical MEC systems are subject to various uncertainties rendering these assumptions impractical. In this paper, we investigate the energy-efficient computation offloading problem by relaxing those common assumptions and considering intrinsic uncertainties in the network. Specifically, we minimize the worst-case expected energy consumption of a local device when executing a time-critical application modeled as a directed acyclic graph. We employ the extreme value theory to bound the occurrence probability of uncertain events. To solve the formulated problem, we develop an $ε$-bounded approximation algorithm based on column generation. The proposed algorithm can efficiently identify a feasible solution that is less than (1+$ε$) of the optimal one. We implement the proposed scheme on an Android smartphone and conduct extensive experiments using a real-world application. Experiment results corroborate that it will lead to lower energy consumption for the client device by considering the intrinsic uncertainties during computation offloading. The proposed computation offloading scheme also significantly outperforms other schemes in terms of energy saving.

cs.DC↗

arXiv2021

HyCA: A Hybrid Computing Architecture for Fault Tolerant Deep Learning

Hardware faults on the regular 2-D computing array of a typical deep learning accelerator (DLA) can lead to dramatic prediction accuracy loss. Prior redundancy design approaches typically have each homogeneous redundant processing element (PE) to mitigate faulty PEs for a limited region of the 2-D computing array rather than the entire computing array to avoid the excessive hardware overhead. However, they fail to recover the computing array when the number of faulty PEs in any region exceeds the number of redundant PEs in the same region. The mismatch problem deteriorates when the fault injection rate rises and the faults are unevenly distributed. To address the problem, we propose a hybrid computing architecture (HyCA) for fault-tolerant DLAs. It has a set of dot-production processing units (DPPUs) to recompute all the operations that are mapped to the faulty PEs despite the faulty PE locations. According to our experiments, HyCA shows significantly higher reliability, scalability, and performance with less chip area penalty when compared to the conventional redundancy approaches. Moreover, by taking advantage of the flexible recomputing, HyCA can also be utilized to scan the entire 2-D computing array and detect the faulty PEs effectively at runtime.

cs.AR↗

arXiv2021

Optimization of Controlled-Z Gate with Data-Driven Gradient Ascent Pulse Engineering in a Superconducting Qubit System

The experimental optimization of a two-qubit controlled-Z (CZ) gate is realized following two different data-driven gradient ascent pulse engineering (GRAPE) protocols in the aim of optimizing the gate operator and the output quantum state, respectively. For both GRAPE protocols, the key computation of gradients utilizes mixed information of the input Z-control pulse and the experimental measurement. With an imperfect initial pulse in a flattop waveform, our experimental implementation shows that the CZ gate is quickly improved and the gate fidelities subject to the two optimized pulses are around 99%. Our experimental study confirms the applicability of the data-driven GRAPE protocols in the problem of the gate optimization.

quant-ph↗