arXiv subjects

Shengyu Zhang

Publications and source records attributed to Shengyu Zhang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Quantum Recurrent Embedding Neural Network

Quantum neural networks have emerged as promising quantum machine learning models, leveraging the properties of quantum systems and classical optimization to solve complex problems in physics and beyond. However, previous studies have demonstrated inevitable trainability issues that severely limit their capabilities in the large-scale regime. In this work, we propose a quantum recurrent embedding neural network (QRENN) inspired by fast-track information pathways in ResNet and general quantum circuit architectures in quantum information theory. By employing dynamical Lie algebras, we provide a rigorous proof of the trainability of QRENN circuits, demonstrating that this deep quantum neural network can avoid barren plateaus. Notably, the general QRENN architecture resists classical simulation as it encompasses powerful quantum circuits such as QSP, QSVT, and DQC1, which are widely believed to be classically intractable. Building on this theoretical foundation, we apply our QRENN to accurately classify quantum Hamiltonians and detect symmetry-protected topological phases, demonstrating its applicability in quantum supervised learning. Our results highlight the power of recurrent data embedding in quantum neural networks and the potential for scalable quantum supervised learning in predicting physical properties and solving complex problems.

quant-ph↗

arXiv2025

Device-Cloud Collaborative Correction for On-Device Recommendation

With the rapid development of recommendation models and device computing power, device-based recommendation has become an important research area due to its better real-time performance and privacy protection. Previously, Transformer-based sequential recommendation models have been widely applied in this field because they outperform Recurrent Neural Network (RNN)-based recommendation models in terms of performance. However, as the length of interaction sequences increases, Transformer-based models introduce significantly more space and computational overhead compared to RNN-based models, posing challenges for device-based recommendation. To balance real-time performance and high performance on devices, we propose Device-Cloud \underline{Co}llaborative \underline{Corr}ection Framework for On-Device \underline{Rec}ommendation (CoCorrRec). CoCorrRec uses a self-correction network (SCN) to correct parameters with extremely low time cost. By updating model parameters during testing based on the input token, it achieves performance comparable to current optimal but more complex Transformer-based models. Furthermore, to prevent SCN from overfitting, we design a global correction network (GCN) that processes hidden states uploaded from devices and provides a global correction solution. Extensive experiments on multiple datasets show that CoCorrRec outperforms existing Transformer-based and RNN-based device recommendation models in terms of performance, with fewer parameters and lower FLOPs, thereby achieving a balance between real-time performance and high efficiency.

cs.IR↗

arXiv2025

MNN-LLM: A Generic Inference Engine for Fast Large Language Model Deployment on Mobile Devices

Large language models (LLMs) have demonstrated exceptional performance across a variety of tasks. However, their substantial scale leads to significant computational resource consumption during inference, resulting in high costs. Consequently, edge device inference presents a promising solution. The primary challenges of edge inference include memory usage and inference speed. This paper introduces MNN-LLM, a framework specifically designed to accelerate the deployment of large language models on mobile devices. MNN-LLM addresses the runtime characteristics of LLMs through model quantization and DRAM-Flash hybrid storage, effectively reducing memory usage. It rearranges weights and inputs based on mobile CPU instruction sets and GPU characteristics while employing strategies such as multicore load balancing, mixed-precision floating-point operations, and geometric computations to enhance performance. Notably, MNN-LLM achieves up to a 8.6x speed increase compared to current mainstream LLM-specific frameworks.

cs.LG↗

arXiv2025

On generating direct powers of dynamical Lie algebras

The expressibility and trainability of parameterized quantum circuits has been shown to be intimately related to their associated dynamical Lie algebras (DLAs). From a quantum algorithm design perspective, given a set $A$ of DLA generators, two natural questions arise: (i) what is the DLA $\mathfrak{g}_{A}$ generated by ${A}$; and (ii) how does modifying the generator set lead to changes in the resulting DLA. While the first question has been the subject of significant attention, much less has been done regarding the second. In this work we focus on the second question, and show how modifying ${A}$ can result in a generator set ${A}'$ such that $\mathfrak{g}_{{A}'}\cong \bigoplus_{j=1}^{K}\mathfrak{g}_{A}$, for some $K \ge 1$. In other words, one generates the direct sum of $K$ copies of the original DLA. In particular, we give qubit- and parameter-efficient ways of achieving this, using only $\log K$ additional qubits, and only a constant factor increase in the number of DLA generators. For cyclic DLAs, which include Pauli DLAs and QAOA-MaxCut DLAs as special cases, this can be done with $\log K $ additional qubits and the same number of DLA generators as ${A}$.

quant-ph↗

arXiv2025

MadaKV: Adaptive Modality-Perception KV Cache Eviction for Efficient Multimodal Long-Context Inference

This paper introduces MadaKV, a modality-adaptive key-value (KV) cache eviction strategy designed to enhance the efficiency of multimodal large language models (MLLMs) in long-context inference. In multimodal scenarios, attention heads exhibit varying preferences for different modalities, resulting in significant disparities in modality importance across attention heads. Traditional KV cache eviction methods, which are tailored for unimodal settings, fail to capture modality-specific information, thereby yielding suboptimal performance. MadaKV addresses these challenges through two key components: modality preference adaptation and hierarchical compression compensation. By dynamically sensing modality information within attention heads and adaptively retaining critical tokens, MadaKV achieves substantial reductions in KV cache memory footprint and model inference decoding latency (1.3 to 1.5 times improvement) while maintaining high accuracy across various multimodal long-context tasks. Extensive experiments on representative MLLMs and the MileBench benchmark demonstrate the effectiveness of MadaKV compared to existing KV cache eviction methods.

cs.LG↗

arXiv2025

Cuff-KT: Tackling Learners' Real-time Learning Pattern Adjustment via Tuning-Free Knowledge State Guided Model Updating

Knowledge Tracing (KT) is a core component of Intelligent Tutoring Systems, modeling learners' knowledge state to predict future performance and provide personalized learning support. Traditional KT models assume that learners' learning abilities remain relatively stable over short periods or change in predictable ways based on prior performance. However, in reality, learners' abilities change irregularly due to factors like cognitive fatigue, motivation, and external stress -- a task introduced, which we refer to as Real-time Learning Pattern Adjustment (RLPA). Existing KT models, when faced with RLPA, lack sufficient adaptability, because they fail to timely account for the dynamic nature of different learners' evolving learning patterns. Current strategies for enhancing adaptability rely on retraining, which leads to significant overfitting and high time overhead issues. To address this, we propose Cuff-KT, comprising a controller and a generator. The controller assigns value scores to learners, while the generator generates personalized parameters for selected learners. Cuff-KT controllably adapts to data changes fast and flexibly without fine-tuning. Experiments on five datasets from different subjects demonstrate that Cuff-KT significantly improves the performance of five KT models with different structures under intra- and inter-learner shifts, with an average relative increase in AUC of 10% and 4%, respectively, at a negligible time cost, effectively tackling RLPA task. Our code and datasets are fully available at https://github.com/zyy-2001/Cuff-KT.

cs.LG↗

arXiv2025

Multimodal LLM-Guided Semantic Correction in Text-to-Image Diffusion

Diffusion models have become the mainstream architecture for text-to-image generation, achieving remarkable progress in visual quality and prompt controllability. However, current inference pipelines generally lack interpretable semantic supervision and correction mechanisms throughout the denoising process. Most existing approaches rely solely on post-hoc scoring of the final image, prompt filtering, or heuristic resampling strategies-making them ineffective in providing actionable guidance for correcting the generative trajectory. As a result, models often suffer from object confusion, spatial errors, inaccurate counts, and missing semantic elements, severely compromising prompt-image alignment and image quality. To tackle these challenges, we propose MLLM Semantic-Corrected Ping-Pong-Ahead Diffusion (PPAD), a novel framework that, for the first time, introduces a Multimodal Large Language Model (MLLM) as a semantic observer during inference. PPAD performs real-time analysis on intermediate generations, identifies latent semantic inconsistencies, and translates feedback into controllable signals that actively guide the remaining denoising steps. The framework supports both inference-only and training-enhanced settings, and performs semantic correction at only extremely few diffusion steps, offering strong generality and scalability. Extensive experiments demonstrate PPAD's significant improvements.

cs.CV↗

arXiv2025

Full Characterization of the Depth Overhead for Quantum Circuit Compilation with Arbitrary Qubit Connectivity Constraint

In some physical implementations of quantum computers, 2-qubit operations can be applied only on certain pairs of qubits. Compilation of a quantum circuit into one compliant to such qubit connectivity constraint results in an increase of circuit depth. Various compilation algorithms were studied, yet what this depth overhead is remains elusive. In this paper, we fully characterize the depth overhead by the routing number of the underlying constraint graph, a graph-theoretic measure which has been studied for 3 decades. We also give reduction algorithms between different graphs, which allow compilation for one graph to be transferred to one for another. These results, when combined with existing routing algorithms, give asymptotically optimal compilation for all commonly seen connectivity graphs in quantum computing.

quant-ph↗

arXiv2025

Depth-Efficient Quantum Circuit Synthesis for Deterministic Dicke State Preparation

The $n$-qubit $k$-weight Dicke states $|D^n_k\rangle$, defined as the uniform superposition of all computational basis states with exactly $k$ qubits in state $|1\rangle$, form a basis of the symmetric subspace and represent an important class of entangled quantum states with broad applications in quantum computing. We propose deterministic quantum circuits for Dicke state preparation under two commonly seen qubit connectivity constraints: 1. All-to-all qubit connectivity: our circuit has depth $O(\log(k)\log(n/k)+k)$, which improves the previous best bound of $O(k\log(n/k))$. 2. Grid qubit connectivity ($(n_1\times n_2)$-grid, $n_1\le n_2$): (a) For $k\ge n_2/n_1$, we design a circuit with depth $O(k\log(n/k)+n_2)$, surpassing the prior $O(\sqrt{nk})$ bound. (b) For $k< n_2/n_1$, we design an optimal-depth circuit with depth $O(n_2)$. Furthermore, we establish the depth lower bounds of $Ω(\log(n))$ for all-to-all qubit connectivity and $Ω(n_2)$ for $(n_1\times n_2)$-grid connectivity constraints, demonstrating the near-optimality of our constructions.

quant-ph↗

arXiv2025

InfiGUI-R1: Advancing Multimodal GUI Agents from Reactive Actors to Deliberative Reasoners

Multimodal Large Language Models (MLLMs) have powered Graphical User Interface (GUI) Agents, showing promise in automating tasks on computing devices. Recent works have begun exploring reasoning in GUI tasks with encouraging results. However, many current approaches rely on manually designed reasoning templates, which may result in reasoning that is not sufficiently robust and adaptive for complex GUI environments. Meanwhile, some existing agents continue to operate as Reactive Actors, relying primarily on implicit reasoning that may lack sufficient depth for GUI tasks demanding planning and error recovery. We argue that advancing these agents requires a shift from reactive acting towards acting based on deliberate reasoning. To facilitate this transformation, we introduce InfiGUI-R1, an MLLM-based GUI agent developed through our Actor2Reasoner framework, a reasoning-centric, two-stage training approach designed to progressively evolve agents from Reactive Actors to Deliberative Reasoners. The first stage, Reasoning Injection, focuses on establishing a basic reasoner. We employ Spatial Reasoning Distillation to transfer cross-modal spatial reasoning capabilities from teacher models to MLLMs through trajectories with explicit reasoning steps, enabling models to integrate GUI visual-spatial information with logical reasoning before action generation. The second stage, Deliberation Enhancement, refines the basic reasoner into a deliberative one using Reinforcement Learning. This stage introduces two approaches: Sub-goal Guidance, which rewards models for generating accurate intermediate sub-goals, and Error Recovery Scenario Construction, which creates failure-and-recovery training scenarios from identified prone-to-error steps. Experimental results show InfiGUI-R1 achieves strong performance in GUI grounding and trajectory tasks. Resources at https://github.com/Reallm-Labs/InfiGUI-R1.

cs.AI↗

arXiv2025

Cascaded Self-Evaluation Augmented Training for Lightweight Multimodal LLMs

Efficient Multimodal Large Language Models (EMLLMs) can improve performance through Chain-of-Thought (CoT) reasoning, but they have poor self-evaluation capabilities during the CoT reasoning process. This is due to their tendency to simplify the reasoning process and the degradation of self-evaluation ability during downstream task fine-tuning. To address this, we intuitively propose \textit{Self-Evaluation Augmented Training (SEAT)}, which uses more powerful EMLLMs to evaluate CoT reasoning data. The evaluation data is then used to train EMLLMs. However, due to the difficulties EMLLMs face with processing long token input-output sequences, and the degradation of self-evaluation ability as a basis for CoT reasoning, the SEAT method is not fully adapted. Therefore, we further propose \textit{Cascaded Self-Evaluation Augmented Training (Cas-SEAT)}, which converts long prompts into cascaded short prompts, each focusing on a specific task. Additionally, we mix CoT reasoning and self-evaluation data to preserve its CoT reasoning ability while enhancing the self-evaluation capability of EMLLMs. We also conduct \textit{Double-level Data Filtering (DDF)}, which includes source data filtering and labeled data filtering, using both manual selection and MLLMs for filtering. Cas-SEAT and DDF work together to improve the performance of EMLLMs. Experiments show that Cas-SEAT achieves an average improvement of 22.16% across multiple datasets, and DDF significantly reduces the resource consumption of training

cs.CL↗

arXiv2025

Disentangled Knowledge Tracing for Alleviating Cognitive Bias

In the realm of Intelligent Tutoring System (ITS), the accurate assessment of students' knowledge states through Knowledge Tracing (KT) is crucial for personalized learning. However, due to data bias, $\textit{i.e.}$, the unbalanced distribution of question groups ($\textit{e.g.}$, concepts), conventional KT models are plagued by cognitive bias, which tends to result in cognitive underload for overperformers and cognitive overload for underperformers. More seriously, this bias is amplified with the exercise recommendations by ITS. After delving into the causal relations in the KT models, we identify the main cause as the confounder effect of students' historical correct rate distribution over question groups on the student representation and prediction score. Towards this end, we propose a Disentangled Knowledge Tracing (DisKT) model, which separately models students' familiar and unfamiliar abilities based on causal effects and eliminates the impact of the confounder in student representation within the model. Additionally, to shield the contradictory psychology ($\textit{e.g.}$, guessing and mistaking) in the students' biased data, DisKT introduces a contradiction attention mechanism. Furthermore, DisKT enhances the interpretability of the model predictions by integrating a variant of Item Response Theory. Experimental results on 11 benchmarks and 3 synthesized datasets with different bias strengths demonstrate that DisKT significantly alleviates cognitive bias and outperforms 16 baselines in evaluation accuracy.

cs.LG↗

arXiv2025

Optimize Incompatible Parameters through Compatibility-aware Knowledge Integration

Deep neural networks have become foundational to advancements in multiple domains, including recommendation systems, natural language processing, and so on. Despite their successes, these models often contain incompatible parameters that can be underutilized or detrimental to model performance, particularly when faced with specific, varying data distributions. Existing research excels in removing such parameters or merging the outputs of multiple different pretrained models. However, the former focuses on efficiency rather than performance, while the latter requires several times more computing and storage resources to support inference. In this paper, we set the goal to explicitly improve these incompatible parameters by leveraging the complementary strengths of different models, thereby directly enhancing the models without any additional parameters. Specifically, we propose Compatibility-aware Knowledge Integration (CKI), which consists of Parameter Compatibility Assessment and Parameter Splicing, which are used to evaluate the knowledge content of multiple models and integrate the knowledge into one model, respectively. The integrated model can be used directly for inference or for further fine-tuning. We conduct extensive experiments on various datasets for recommendation and language tasks, and the results show that Compatibility-aware Knowledge Integration can effectively optimize incompatible parameters under multiple tasks and settings to break through the training limit of the original model without increasing the inference cost.

cs.LG↗

arXiv2025

Collaboration of Large Language Models and Small Recommendation Models for Device-Cloud Recommendation

Large Language Models (LLMs) for Recommendation (LLM4Rec) is a promising research direction that has demonstrated exceptional performance in this field. However, its inability to capture real-time user preferences greatly limits the practical application of LLM4Rec because (i) LLMs are costly to train and infer frequently, and (ii) LLMs struggle to access real-time data (its large number of parameters poses an obstacle to deployment on devices). Fortunately, small recommendation models (SRMs) can effectively supplement these shortcomings of LLM4Rec diagrams by consuming minimal resources for frequent training and inference, and by conveniently accessing real-time data on devices. In light of this, we designed the Device-Cloud LLM-SRM Collaborative Recommendation Framework (LSC4Rec) under a device-cloud collaboration setting. LSC4Rec aims to integrate the advantages of both LLMs and SRMs, as well as the benefits of cloud and edge computing, achieving a complementary synergy. We enhance the practicability of LSC4Rec by designing three strategies: collaborative training, collaborative inference, and intelligent request. During training, LLM generates candidate lists to enhance the ranking ability of SRM in collaborative scenarios and enables SRM to update adaptively to capture real-time user interests. During inference, LLM and SRM are deployed on the cloud and on the device, respectively. LLM generates candidate lists and initial ranking results based on user behavior, and SRM get reranking results based on the candidate list, with final results integrating both LLM's and SRM's scores. The device determines whether a new candidate list is needed by comparing the consistency of the LLM's and SRM's sorted lists. Our comprehensive and extensive experimental analysis validates the effectiveness of each strategy in LSC4Rec.

cs.IR↗

arXiv2025

Reinforcement Learning Enhanced LLMs: A Survey

Reinforcement learning (RL) enhanced large language models (LLMs), particularly exemplified by DeepSeek-R1, have exhibited outstanding performance. Despite the effectiveness in improving LLM capabilities, its implementation remains highly complex, requiring complex algorithms, reward modeling strategies, and optimization techniques. This complexity poses challenges for researchers and practitioners in developing a systematic understanding of RL-enhanced LLMs. Moreover, the absence of a comprehensive survey summarizing existing research on RL-enhanced LLMs has limited progress in this domain, hindering further advancements. In this work, we are going to make a systematic review of the most up-to-date state of knowledge on RL-enhanced LLMs, attempting to consolidate and analyze the rapidly growing research in this field, helping researchers understand the current challenges and advancements. Specifically, we (1) detail the basics of RL; (2) introduce popular RL-enhanced LLMs; (3) review researches on two widely-used reward model-based RL techniques: Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF); and (4) explore Direct Preference Optimization (DPO), a set of methods that bypass the reward model to directly use human preference data for aligning LLM outputs with human expectations. We will also point out current challenges and deficiencies of existing methods and suggest some avenues for further improvements. Project page of this work can be found at https://github.com/ShuheWang1998/Reinforcement-Learning-Enhanced-LLMs-A-Survey.

cs.CL↗

arXiv2025

InfiFusion: A Unified Framework for Enhanced Cross-Model Reasoning via LLM Fusion

We introduce InfiFusion, an efficient training pipeline designed to integrate multiple domain-specialized Large Language Models (LLMs) into a single pivot model, effectively harnessing the strengths of each source model. Traditional fusion methods either merge model parameters directly or rely on knowledge distillation with rigid assumptions, limiting their flexibility and efficiency. InfiFusion overcomes these limitations by enhancing Universal Logit Distillation (ULD) with Top-K selection and Logits Standardization. We propose two fusion strategies: Pairwise Fusion (InfiFusion$_p$), where each source model knowledge is distilled individually into the pivot model followed by merging and Unified Fusion (InfiFusion$_u$), where knowledge from all source models is distilled simultaneously into the pivot model. InfiFusion outperforms the state-of-the-art models, such as Qwen-2.5-14B-Instruct and Phi-4, across 11 widely applied benchmarks covering reasoning, coding, mathematics, and instruction-following tasks. Notably, InfiFusion achieves this superior performance while significantly reduces computational costs, completing full training with only 160 H800 GPU hours compared to the millions typically required for traditional LLM training.

cs.CL↗

arXiv2025

InfiR : Crafting Effective Small Language Models and Multimodal Small Language Models in Reasoning

Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have made significant advancements in reasoning capabilities. However, they still face challenges such as high computational demands and privacy concerns. This paper focuses on developing efficient Small Language Models (SLMs) and Multimodal Small Language Models (MSLMs) that retain competitive reasoning abilities. We introduce a novel training pipeline that enhances reasoning capabilities and facilitates deployment on edge devices, achieving state-of-the-art performance while minimizing development costs. \InfR~ aims to advance AI systems by improving reasoning, reducing adoption barriers, and addressing privacy concerns through smaller model sizes. Resources are available at https://github. com/Reallm-Labs/InfiR.

cs.CL↗

arXiv2025

Quantum Approximate Optimization Algorithms for Maximum Cut on Low-Girth Graphs

Maximum cut (MaxCut) on graphs is a classic NP-hard problem. In quantum computing, Farhi, Gutmann, and Goldstone proposed the Quantum Approximate Optimization Algorithm (QAOA) for solving the MaxCut problem. Its guarantee on cut fraction (the fraction of edges in the output cut over all edges) was mainly studied for high-girth graphs, i.e., graphs with only long cycles. On the other hand, low-girth graphs are ubiquitous in theoretical computer science, including expander graphs being outstanding examples with wide applications in theory and beyond. In this paper, we apply QAOA to MaxCut on a set of expander graphs proposed by Mohanty and O'Donnell known as additive product graphs. Additionally, we apply multi-angle QAOA (ma-QAOA) to better utilize the graph structure of additive product graphs in ansatz design. In theory, we derive an iterative formula to calculate the expected cut fraction of such graphs. This formula also extends to the quantum MaxCut problem. On the other hand, we conduct numerical experiments to compare between best-known classical local algorithms and QAOA with constant depth. Our results demonstrate that QAOA outperforms the best-known classical algorithms by 0.3% to 5.2% on several additive product graphs, while ma-QAOA further enhances this advantage by an additional 0.6% to 2.5%. In particular, we observe cases that ma-QAOA exhibits superiority over best-known classical algorithms but QAOA does not. Furthermore, we extend our experiments to planar graphs such as tiling grid graphs, where QAOA also demonstrates an advantage.

quant-ph↗