arXiv subjects

Dong Li

Publications and source records attributed to Dong Li.

At least 91 records · Page 5Linked to original sources

arXiv2026

HybridGen: Efficient LLM Generative Inference via CPU-GPU Hybrid Computing

As modern LLMs support thousands to millions of tokens, KV caches grow to hundreds of gigabytes, stressing memory capacity and bandwidth. Existing solutions, such as KV cache pruning and offloading, alleviate these but underutilize hardware by relying solely on either GPU or CPU for attention computing, and considering yet limited CPU local memory for KV cache storage. We propose HybridGen, an efficient hybrid attention framework for long-context LLM inference. HybridGen enables CPU-GPU collaborative attention on systems with expanded tiered memory (e.g., CXL memory), addressing three key challenges: (1) multi-dimensional attention dependencies, (2) intensifying CPU-GPU load imbalance with longer sequences, and (3) NUMA penalty of tiered memories. HybridGen tackles these by introducing attention logit parallelism, a feedback-driven scheduler, and semantic-aware KV cache mapping. Experiments with three LLM models with eleven different sizes on three GPU platforms with a CXL-expanded memory show that HybridGen outperforms six state-of-the-art KV cache management methods by 1.41x--3.2x on average while maintaining superior accuracy.

cs.PF↗

arXiv2026

When to Trust Tools? Adaptive Tool Trust Calibration For Tool-Integrated Math Reasoning

Large reasoning models (LRMs) have achieved strong performance enhancement through scaling test time computation, but due to the inherent limitations of the underlying language models, they still have shortcomings in tasks that require precise computation and extensive knowledge reserves. Tool-Integrated Reasoning (TIR) has emerged as a promising paradigm that incorporates tool call and execution within the reasoning trajectory. Although recent works have released some powerful open-source TIR models, our analysis reveals that these models still suffer from critical deficiencies. We find that when the reasoning of the model conflicts with the tool results, the model tends to believe in its own reasoning. And there are cases where the tool results are correct but are ignored by the model, resulting in incorrect answers, which we define as "Tool Ignored''. This indicates that the model does not know when to trust or ignore the tool. To overcome these limitations, We introduce Adaptive Tool Trust Calibration (ATTC), a novel framework that guides the model to adaptively choose to trust or ignore the tool results based on the confidence score of generated code blocks. The experimental results from various open-source TIR models of different sizes and across multiple datasets demonstrate that ATTC effectively reduces the "Tool Ignored" issue, resulting in a performance increase of 4.1% to 7.5%.

cs.CL↗

arXiv2026

Bird-SR: Bidirectional Reward-Guided Diffusion for Real-World Image Super-Resolution

Powered by multimodal text-to-image priors, diffusion-based super-resolution excels at synthesizing intricate details; however, models trained on synthetic low-resolution (LR) and high-resolution (HR) image pairs often degrade when applied to real-world LR images due to significant distribution shifts. We propose Bird-SR, a bidirectional reward-guided diffusion framework that formulates super-resolution as trajectory-level preference optimization via reward feedback learning (ReFL), jointly leveraging synthetic LR-HR pairs and real-world LR images. For structural fidelity easily affected in ReFL, the model is directly optimized on synthetic pairs at early diffusion steps, which also facilitates structure preservation for real-world inputs under smaller distribution gap in structure levels. For perceptual enhancement, quality-guided rewards are applied to both synthetic and real LR images at the later trajectory phase. To mitigate reward hacking, the rewards for synthetic results are formulated in a relative advantage space bounded by their ground-truth counterparts, while real-world optimization is regularized via a semantic alignment constraint. Furthermore, to balance structural and perceptual learning, we introduce a dynamic fidelity-perception weighting strategy that emphasizes structure preservation at early stages and progressively shifts focus toward perceptual optimization at later diffusion steps. Extensive experiments on real-world SR benchmarks demonstrate that Bird-SR consistently outperforms state-of-the-art methods in perceptual quality while preserving structural consistency, validating its effectiveness for real-world super-resolution. Our code can be obtained at https://github.com/fanzh03/Bird-SR.

cs.CV↗

arXiv2026

Hybrid Adaptive Tuning for Tiered Memory Systems

Memory tiering provides a cost-effective solution to increase memory capacity, utilization, and even bandwidth. Memory tiering relies on system software for memory profiling, detection of frequently accessed pages, and page migration. Such a system software often comes with system parameters. The configurations of those parameters impact application performance. We comprehensively classify system parameters, and characterize the sensitivity of application performance to them using representative memory tiering solutions. Furthermore, we introduce a lightweight and user-friendly framework PTMT, which automates tuning of parameters at runtime for various memory tiering solutions. We identify major challenges for online tuning of memory tiering. PTMT uses a hybrid "offline + online" tuning method: while the offline phase builds a performance database for online queries and reduces runtime overhead, the online phase uses reinforcement learning (customized to memory tiering) to tune. PTMT improves performance by 30%, 26%, 21%, and 14%, on four memory tiering solutions (TPP, UPM, Colloid, and AutoNUMA), compared to using the default configurations. PTMT outperforms the state-of-the-art by 32% on average.

cs.OS↗

arXiv2026

Iterative Inference-time Scaling with Adaptive Frequency Steering for Image Super-Resolution

Diffusion models have become a leading paradigm for image super-resolution (SR), but existing methods struggle to guarantee both the high-frequency perceptual quality and the low-frequency structural fidelity of generated images. Although inference-time scaling can theoretically improve this trade-off by allocating more computation, existing strategies remain suboptimal: reward-driven particle optimization often causes perceptual over-smoothing, while optimal-path search tends to lose structural consistency. To overcome these difficulties, we propose Iterative Diffusion Inference-Time Scaling with Adaptive Frequency Steering (IAFS), a training-free framework that jointly leverages iterative refinement and frequency-aware particle fusion. IAFS addresses the challenge of balancing perceptual quality and structural fidelity by progressively refining the generated image through iterative correction of structural deviations. Simultaneously, it ensures effective frequency fusion by adaptively integrating high-frequency perceptual cues with low-frequency structural information, allowing for a more accurate and balanced reconstruction across different image details. Extensive experiments across multiple diffusion-based SR models show that IAFS effectively resolves the perception-fidelity conflict, yielding consistently improved perceptual detail and structural accuracy, and outperforming existing inference-time scaling methods.

cs.CV↗

arXiv2026

Channel-Aware Preemptive Scheduling for Semantic Communication with Truncated Diffusion and Path Compensation

Semantic communication (SemCom) presents a transformative paradigm for alleviating bandwidth limitations in mobile networks by transmitting task-relevant semantic features instead of raw data bits. While SemCom systems utilizing diffusion models achieve superior generation quality, existing research treats semantic generation and wireless transmission as temporally independent processes. This separation neglects the intrinsic conflict between the multi-step iterative delays inherent in diffusion models and the time-varying fading characteristics of wireless channels. To address this discrepancy, this paper proposes a channel-aware preemptive scheduling with truncated diffusion and path compensation (CAPS-TDPC) framework. Contrary to conventional methods that require completion of the generation phase prior to transmission, the proposed framework implements a channel-driven scheduling mechanism: each user maintains a countdown inversely proportional to its instantaneous channel gain, and the user with the shortest countdown transmits immediately, regardless of whether its diffusion process has completed. This design permits the interruption of the forward diffusion process to enable early transmission under favorable channel conditions. In addition, a receiver-side compensation mechanism grounded in path dynamics is developed to mitigate the semantic loss resulting from such interruptions. A path deficit metric is proposed at the receiver to quantify the recovery difficulty of distinct image blocks by incorporating the velocity field of the inverse dynamics model, which allows for adaptive weighted inverse sampling. Experimental evaluations demonstrate that the proposed framework substantially reduces the end-to-end latency while maintaining the high-fidelity semantic reconstruction, thereby enhancing the system robustness in fast fading channel environments.

cs.IT↗

arXiv2026

DeepGuard: Secure Code Generation via Multi-Layer Semantic Aggregation

Large Language Models (LLMs) for code generation can replicate insecure patterns from their training data. To mitigate this, a common strategy for security hardening is to fine-tune models using supervision derived from the final transformer layer. However, this design may suffer from a final-layer bottleneck: vulnerability-discriminative cues can be distributed across layers and become less detectable near the output representations optimized for next-token prediction. To diagnose this issue, we perform layer-wise linear probing. We observe that vulnerability-related signals are most detectable in a band of intermediate-to-upper layers yet attenuate toward the final layers. Motivated by this observation, we introduce DeepGuard, a framework that leverages distributed security-relevant cues by aggregating representations from multiple upper layers via an attention-based module. The aggregated signal powers a dedicated security analyzer within a multi-objective training objective that balances security enhancement and functional correctness, and further supports a lightweight inference-time steering strategy. Extensive experiments across five code LLMs demonstrate that DeepGuard improves the secure-and-correct generation rate by an average of 11.9% over strong baselines such as SVEN. It also preserves functional correctness while exhibiting generalization to held-out vulnerability types. Our code is public at https://github.com/unknownhl/DeepGuard.

cs.SE↗

arXiv2026

ForgeryGPT: A Multimodal LLM for Interpretable Image Forgery Detection and Localization

Multimodal Large Language Models (MLLMs), such as GPT4o, have shown strong capabilities in visual reasoning and explanation generation. However, despite these strengths, they face significant challenges in the increasingly critical task of Image Forgery Detection and Localization (IFDL). Moreover, existing IFDL methods are typically limited to the learning of low-level semantic-agnostic clues and merely provide a single outcome judgment. To tackle these issues, we propose ForgeryGPT, a novel framework that advances the IFDL task by capturing high-order forensics knowledge correlations of forged images from diverse linguistic feature spaces, while enabling explainable generation and interactive dialogue through a newly customized Large Language Model (LLM) architecture. Specifically, ForgeryGPT enhances traditional LLMs by integrating the Mask-Aware Forgery Extractor, which enables the excavating of precise forgery mask information from input images and facilitating pixel-level understanding of tampering artifacts. The Mask-Aware Forgery Extractor consists of a Forgery Localization Expert (FL-Expert) and a Mask Encoder, where the FL-Expert is augmented with an Object-agnostic Forgery Prompt and a Vocabulary-enhanced Vision Encoder, allowing for effectively capturing of multi-scale fine-grained forgery details. To enhance its performance, we implement a three-stage training strategy, supported by our designed Mask-Text Alignment and IFDL Task-Specific Instruction Tuning datasets, which align vision-language modalities and improve forgery detection and instruction-following capabilities. Extensive experiments demonstrate the effectiveness of the proposed method.

cs.CV↗

arXiv2026

FinPercep-RM: A Fine-grained Reward Model and Co-evolutionary Curriculum for RL-based Real-world Super-Resolution

Reinforcement Learning with Human Feedback (RLHF) has proven effective in image generation field guided by reward models to align human preferences. Motivated by this, adapting RLHF for Image Super-Resolution (ISR) tasks has shown promise in optimizing perceptual quality with Image Quality Assessment (IQA) model as reward models. However, the traditional IQA model usually output a single global score, which are exceptionally insensitive to local and fine-grained distortions. This insensitivity allows ISR models to produce perceptually undesirable artifacts that yield spurious high scores, misaligning optimization objectives with perceptual quality and results in reward hacking. To address this, we propose a Fine-grained Perceptual Reward Model (FinPercep-RM) based on an Encoder-Decoder architecture. While providing a global quality score, it also generates a Perceptual Degradation Map that spatially localizes and quantifies local defects. We specifically introduce the FGR-30k dataset to train this model, consisting of diverse and subtle distortions from real-world super-resolution models. Despite the success of the FinPercep-RM model, its complexity introduces significant challenges in generator policy learning, leading to training instability. To address this, we propose a Co-evolutionary Curriculum Learning (CCL) mechanism, where both the reward model and the ISR model undergo synchronized curricula. The reward model progressively increases in complexity, while the ISR model starts with a simpler global reward for rapid convergence, gradually transitioning to the more complex model outputs. This easy-to-hard strategy enables stable training while suppressing reward hacking. Experiments validates the effectiveness of our method across ISR models in both global quality and local realism on RLHF methods.

cs.CV↗

arXiv2026

Ultrafast Non-Volatile Weyl LuminoMem for Mid-Infrared In-Memory Computing

Integrated optoelectronic systems strive to combine the logic/memory density of electronics with the bandwidth of photonics, but monolithic realization is impeded by the inefficient electronic-to-photonic interface. Current architectures rely on separate readout circuitry and modulators, creating bottlenecks in energy and latency, while existing direct transduction methods often compromise on switching speed or non-volatility. Here, we report an ultrafast, non-volatile optoelectronic memory, named LuminoMem, that integrates electrical storage and mid-infrared light emission in a single device. The device utilizes a floating-gate architecture, in which the Weyl semiconductor tellurium serves simultaneously as a charge-trapping storage layer and an emissive medium. This design enables nanosecond-scale electrical programming of non-volatile photoluminescence at 3.4 um, allowing direct optical access to stored states without external modulation. We demonstrate 4-bit (16-level) optical storage capacity and validate the device's performance through neural network simulations that achieve high accuracy on the Fashion-MNIST dataset. By effectively bridging the gap between electronic storage and mid-infrared photonics, the demonstrated mid-infrared LuminoMem provides a hardware foundation for promoting current computation efficiency and potential intelligent platforms that co-integrate computing, memory, and sensing capabilities.

cond-mat.mtrl-sci↗

arXiv2026

CTD-Diff: Cooperative Time-Division Diffusion for Multi-User Semantic Communication Systems

Semantic communication (SemCom) has emerged as a transformative paradigm for efficient information transmission by emphasizing the exchange of task-relevant meaning rather than raw data. While diffusion-based SemCom models have demonstrated remarkable generative capabilities, existing studies predominantly focus on point-to-point links, overlooking the potential of multi-user (MU) cooperation in MU wireless environments. To address this limitation, we propose a Cooperative Time-Division Diffusion (CTD-Diff) framework. Unlike traditional approaches that view channel noise solely as a detriment, our framework innovatively integrates the noisy wireless transmission process directly into the forward diffusion chain. Specifically, we design a multi-user cooperation mechanism based on Time-Division Multiple Access (TDMA), where idle users overhearing the active transmitter act as semantic collaborators. To maximize the signal fidelity, the receiver employs direct signal aggregation to fuse the direct signal with cooperative copies. This aggregated noisy semantic representation serves as the condition for the reverse diffusion process, allowing the receiver to reconstruct high-fidelity data by mitigating the cumulative channel distortions. By effectively converting physical channel noise into diffusion noise, the proposed method significantly enhances the transmission reliability. Extensive experiments demonstrate that CTD-Diff outperforms various baselines regarding the reconstruction accuracy and the perceptual quality, particularly under challenging low signal-to-noise ratio (SNR) conditions.

cs.IT↗

arXiv2026

DiffSparse: Accelerating Diffusion Transformers with Learned Token Sparsity

Diffusion models demonstrate outstanding performance in image generation, but their multi-step inference mechanism requires immense computational cost. Previous works accelerate inference by leveraging layer or token cache techniques to reduce computational cost. However, these methods fail to achieve superior acceleration performance in few-step diffusion transformer models due to inefficient feature caching strategies, manually designed sparsity allocation, and the practice of retaining complete forward computations in several steps in these token cache methods. To tackle these challenges, we propose a differentiable layer-wise sparsity optimization framework for diffusion transformer models, leveraging token caching to reduce token computation costs and enhance acceleration. Our method optimizes layer-wise sparsity allocation in an end-to-end manner through a learnable network combined with a dynamic programming solver. Additionally, our proposed two-stage training strategy eliminates the need for full-step processing in existing methods, further improving efficiency. We conducted extensive experiments on a range of diffusion-transformer models, including DiT-XL/2, PixArt-$α$, FLUX, and Wan2.1. Across these architectures, our method consistently improves efficiency without degrading sample quality. For example, on PixArt-$α$ with 20 sampling steps, we reduce computational cost by $54\%$ while achieving generation metrics that surpass those of the original model, substantially outperforming prior approaches. These results demonstrate that our method delivers large efficiency gains while often improving generation quality.

cs.CV↗

arXiv2026

Closeby Habitable Exoplanet Survey (CHES). V. Planetary Parameters Derived from Angular Separation Variations

The Closeby Habitable Exoplanet Survey (CHES) aims to achieve microarcsecond-level astrometry of about one hundred nearby FGK-type stars within 10 parsecs to detect Earth-like planets. Such precision exceeds the capability of absolute astrometry relying on Gaia catalogs, whose positional accuracy degrades over time due to error propagation from stellar motion and epoch offsets, limiting their use in microarcsecond-level detection. Traditional relative astrometry depends on positional components along right ascension and declination, requiring precise knowledge of field rotation and satellite attitude, which introduces additional errors. To address this, we propose a new relative measurement model based solely on variations in the length of angular separation between the target and reference stars, independent of direction. The model incorporates effects such as proper motion, parallax, radial velocity, light aberration, gravitational lensing, and planetary perturbations, enabling reconstruction of planetary orbits and masses. This approach enhances measurement stability and precision, providing a framework that is not entirely dependent on the Gaia catalog and suitable for CHES and other future high-accuracy astrometric missions.

astro-ph.EP↗

arXiv2026

Effective Rank Analysis and Optimization of Flexible Antenna-Enabled Wireless Systems: Movable Antennas or Pinching Antennas?

Flexible antenna technology has recently emerged as a key enabler for next-generation wireless communications, which can effectively exploit the spatial degrees of freedom (DoF). However, existing conventional metrics (e.g., spectral and energy efficiency) cannot directly measure the variability for different flexible antenna structures for the spatial DoF. To objectively analyze and compare the spatial DoF of different flexible antennas, the effective rank is introduced as a metric for two major flexible antenna technologies, i.e., movable antenna (MA) and pinching antenna (PA)-enabled wireless systems, optimizing their antenna positions over multi-time slots. However, the inherent non-convexity and high computational complexity of the resulting effective rank maximization problems in MA and PA systems render them hard to solve. To circumvent these problems, we propose the graph attention implicit quantile network (GAIQN) and multi-agent graph attention Q-network (MAGAQN) algorithms to enhance the effective rank of MA and PA systems through antenna positioning, respectively. Meanwhile, the novel top-k action selection methods are designed to ensure collision-free between multiple MAs on the two-dimensional array plane or multiple PAs on the same waveguide. Simulation results validate the effectiveness and advancement of our proposed GAIQN and MAGAQN algorithms compared with benchmarks, which enhance the effective rank by at least 1.6% and 1.3%, respectively, while consistently ensuring collision-free between flexible antennas. Besides, under the same number of flexible antennas, the MA system supports a higher effective rank than the PA system, whereas the PA system offers greater stability in terms of achievable spatial DoF.

cs.IT↗

arXiv2026

LLM-Enhanced Energy Contrastive Learning for Out-of-Distribution Detection in Text-Attributed Graphs

Text-attributed graphs, where nodes are enriched with textual attributes, have become a powerful tool for modeling real-world networks such as citation, social, and transaction networks. However, existing methods for learning from these graphs often assume that the distributions of training and testing data are consistent. This assumption leads to significant performance degradation when faced with out-of-distribution (OOD) data. In this paper, we address the challenge of node-level OOD detection in text-attributed graphs, with the goal of maintaining accurate node classification while simultaneously identifying OOD nodes. We propose a novel approach, LLM-Enhanced Energy Contrastive Learning for Out-of-Distribution Detection in Text-Attributed Graphs (LECT), which integrates large language models (LLMs) and energy-based contrastive learning. The proposed method involves generating high-quality OOD samples by leveraging the semantic understanding and contextual knowledge of LLMs to create dependency-aware pseudo-OOD nodes, and applying contrastive learning based on energy functions to distinguish between in-distribution (IND) and OOD nodes. The effectiveness of our method is demonstrated through extensive experiments on six benchmark datasets, where our method consistently outperforms state-of-the-art baselines, achieving both high classification accuracy and robust OOD detection capabilities.

cs.AI↗

arXiv2026

MARLIN: Multi-Agent Reinforcement Learning for Incremental DAG Discovery

Uncovering causal structures from observational data is crucial for understanding complex systems and making informed decisions. While reinforcement learning (RL) has shown promise in identifying these structures in the form of a directed acyclic graph (DAG), existing methods often lack efficiency, making them unsuitable for online applications. In this paper, we propose MARLIN, an efficient multi agent RL based approach for incremental DAG learning. MARLIN uses a DAG generation policy that maps a continuous real valued space to the DAG space as an intra batch strategy, then incorporates two RL agents state specific and state invariant to uncover causal relationships and integrates these agents into an incremental learning framework. Furthermore, the framework leverages a factored action space to enhance parallelization efficiency. Extensive experiments on synthetic and real datasets demonstrate that MARLIN outperforms state of the art methods in terms of both efficiency and effectiveness.

cs.LG↗

arXiv2026

Characterizing the 3D evolution of two successive CMEs heading for Mercury

We studied two successive coronal mass ejections (CMEs) that erupted from the same active region (AR 12994) on 2022 April 15 and propagated toward Mercury. Using multi-view observations, we applied the revised cone model to determine the three-dimensional geometry and the early kinematics of the two CMEs. Our best fit parameters indicate large angular extents of 84 and 86 and propagation directions of 119.0 and 110.4 (measured from the Sun Earth line) for CME1 and CME2, respectively, while that of Mercury is 120.1. The derived axis inclinations are 28 for CME1 and 21 for CME2, consistent with the orientation of the erupting flux ropes in the source region. Height time analysis indicates approximately uniform motion speeds of 636 for CME1 and 696 for CME2, respectively. This paper provides valuable insights for predicting the impact of CMEs heading for Mercury as well as other solar planets in the future.

astro-ph.SR↗

arXiv2026

$\textbf{Re}^{2}$: Unlocking LLM Reasoning via Reinforcement Learning with Re-solving

Reinforcement learning with verifiable rewards (RLVR) has shown promise in enhancing the reasoning performance of large language models (LLMs) by increasing test-time compute. However, even after extensive RLVR training, such models still tend to generate unnecessary and low-quality steps in their chain-of-thought (CoT), leading to inefficient overthinking and lower answer quality. We show that when the initial direction or quality of the CoT is suboptimal, the model often fails to reach the correct answer, even after generating several times more tokens than when the initial CoT is well-initialized. To this end, we introduce Reinforcement Learning with Re-solving (Re$^2$), in which LLMs learn to flexibly abandon unproductive reasoning paths and restart the solution process when necessary, rather than always committing to a final answer. Re$^2$ applies pure reinforcement learning without any preliminary supervised fine-tuning, successfully amplifying the rare redo behavior in vanilla models from only 0.5% to over 30%. This leads to substantial performance gains over standard RLVR under the same training compute budget, and also demonstrates notable improvements in test-time performance as the number of samples increases.

cs.AI↗