arXiv subjects

Gang Li

Publications and source records attributed to Gang Li.

At least 145 records · Page 8Linked to original sources

arXiv2025

Towards Efficient and Accurate Spiking Neural Networks via Adaptive Bit Allocation

Multi-bit spiking neural networks (SNNs) have recently become a heated research spot, pursuing energy-efficient and high-accurate AI. However, with more bits involved, the associated memory and computation demands escalate to the point where the performance improvements become disproportionate. Based on the insight that different layers demonstrate different importance and extra bits could be wasted and interfering, this paper presents an adaptive bit allocation strategy for direct-trained SNNs, achieving fine-grained layer-wise allocation of memory and computation resources. Thus, SNN's efficiency and accuracy can be improved. Specifically, we parametrize the temporal lengths and the bit widths of weights and spikes, and make them learnable and controllable through gradients. To address the challenges caused by changeable bit widths and temporal lengths, we propose the refined spiking neuron, which can handle different temporal lengths, enable the derivation of gradients for temporal lengths, and suit spike quantization better. In addition, we theoretically formulate the step-size mismatch problem of learnable bit widths, which may incur severe quantization errors to SNN, and accordingly propose the step-size renewal mechanism to alleviate this issue. Experiments on various datasets, including the static CIFAR and ImageNet datasets and the dynamic CIFAR-DVS and DVS-GESTURE datasets, demonstrate that our methods can reduce the overall memory and computation cost while achieving higher accuracy. Particularly, our SEWResNet-34 can achieve a 2.69\% accuracy gain and 4.16$\times$ lower bit budgets over the advanced baseline work on ImageNet. This work is open-sourced at \href{https://github.com/Ikarosy/Towards-Efficient-and-Accurate-Spiking-Neural-Networks-via-Adaptive-Bit-Allocation}{this link}.

cs.NE↗

arXiv2025

FedCure: Mitigating Participation Bias in Semi-Asynchronous Federated Learning with Non-IID Data

While semi-asynchronous federated learning (SAFL) combines the efficiency of synchronous training with the flexibility of asynchronous updates, it inherently suffers from participation bias, which is further exacerbated by non-IID data distributions. More importantly, hierarchical architecture shifts participation from individual clients to client groups, thereby further intensifying this issue. Despite notable advancements in SAFL research, most existing works still focus on conventional cloud-end architectures while largely overlooking the critical impact of non-IID data on scheduling across the cloud-edge-client hierarchy. To tackle these challenges, we propose FedCure, an innovative semi-asynchronous Federated learning framework that leverages coalition construction and participation-aware scheduling to mitigate participation bias with non-IID data. Specifically, FedCure operates through three key rules: (1) a preference rule that optimizes coalition formation by maximizing collective benefits and establishing theoretically stable partitions to reduce non-IID-induced performance degradation; (2) a scheduling rule that integrates the virtual queue technique with Bayesian-estimated coalition dynamics, mitigating efficiency loss while ensuring mean rate stability; and (3) a resource allocation rule that enhances computational efficiency by optimizing client CPU frequencies based on estimated coalition dynamics while satisfying delay requirements. Comprehensive experiments on four real-world datasets demonstrate that FedCure improves accuracy by up to 5.1x compared with four state-of-the-art baselines, while significantly enhancing efficiency with the lowest coefficient of variation 0.0223 for per-round latency and maintaining long-term balance across diverse scenarios.

cs.LG↗

arXiv2025

ChemFM as a Scaling Law Guided Foundation Model Pre-trained on Informative Chemicals

Traditional AI methods often rely on task-specific model designs and training, which constrain both the scalability of model size and generalization across different tasks. Here, we introduce ChemFM, a large foundation model specifically developed for chemicals. By conducting a series of scaling experiments, we identify UniChem as the informative molecular database for pre-training the foundation model. ChemFM comprises 3 billion parameters and is pre-trained on 178 million molecules using self-supervised causal language modeling to extract generalizable molecular representations. This model can be adapted to diverse downstream chemical applications using either full-parameter or parameter-efficient fine-tuning methods. ChemFM consistently outperforms state-of-the-art task-specific AI models across all tested tasks. Notably, it achieves up to 67.48% performance improvement across 34 property prediction benchmarks, up to 33.80% reduction in mean average deviation between conditioned and actual properties of generated molecules in conditional molecular generation tasks, and up to 3.7% top-1 accuracy improvement across 4 reaction prediction datasets. Moreover, ChemFM demonstrates its superior performance in predicting antibiotic activity and cytotoxicity, highlighting its potential to advance the discovery of novel antibiotics. Furthermore, we demonstrate that, as a foundation model, ChemFM exhibits strong data efficiency, requiring significantly fewer labeled training samples to achieve state-of-the-art performance. We anticipate that ChemFM will significantly advance chemistry research by providing a foundation model capable of effectively generalizing across a broad range of tasks with minimal additional training.

cs.CE↗

arXiv2025

FLYINGTRUST: A Benchmark for Quadrotor Navigation Across Scenarios and Vehicles

Visual navigation algorithms for quadrotors often exhibit a large variation in performance when transferred across different vehicle platforms and scene geometries, which increases the cost and risk of field deployment. To support systematic early-stage evaluation, we introduce FLYINGTRUST, a high-fidelity, configurable benchmarking framework that measures how platform kinodynamics and scenario structure jointly affect navigation robustness. FLYINGTRUST models vehicle capability with two compact, physically interpretable indicators: maximum thrust-to-weight ratio and axis-wise maximum angular acceleration. The benchmark pairs a diverse scenario library with a heterogeneous set of real and virtual platforms and prescribes a standardized evaluation protocol together with a composite scoring method that balances scenario importance, platform importance and performance stability. We use FLYINGTRUST to compare representative optimization-based and learning-based navigation approaches under identical conditions, performing repeated trials per platform-scenario combination and reporting uncertainty-aware metrics. The results reveal systematic patterns: navigation success depends predictably on platform capability and scene geometry, and different algorithms exhibit distinct preferences and failure modes across the evaluated conditions. These observations highlight the practical necessity of incorporating both platform capability and scenario structure into algorithm design, evaluation, and selection, and they motivate future work on methods that remain robust across diverse platforms and scenarios.

cs.RO↗

arXiv2025

CUARewardBench: A Benchmark for Evaluating Reward Models on Computer-using Agent

Computer-using agents (CUAs) enable task completion through natural interaction with operating systems and software interfaces. While script-based verifiers are widely adopted for evaluation, they suffer from limited scalability and inability to provide step-wise assessment. Reward models offer promising alternatives, but their effectiveness on CUA evaluation remains largely underexplored. To address this gap, we present CUARewardBench, comprising four key contributions: (1) First-ever Comprehensive CUA Reward Benchmark: We introduce the first benchmark for evaluating both outcome reward models (ORM) and process reward models (PRM) on CUA tasks, enabling systematic assessment across trajectory-level and step-level evaluation. (2) Diverse, Practical and Reliable Dataset: CUARewardBench encompasses trajectories from 10 software categories and 7 agent architectures with varying performance levels (25.9%-50.8% success rates). All trajectories are expertly annotated through carefully designed protocols, with rigorous quality control to ensure reliability and practical applicability. (3) Comprehensive Analysis and Insights: Through extensive experiments across 7 vision-language models and 3 prompt templates, we reveal critical limitations of current CUA RMs, including insufficient visual reasoning capabilities, knowledge deficiencies, and the superiority of general VLMs over specialized CUA models for reward evaluation. (4) Unanimous Prompt Ensemble (UPE): Based on the insights from our comprehensive analysis, we propose UPE, a novel ensemble method that significantly enhances reward model reliability through strict unanimous voting and strategic prompt-template configurations. UPE achieves 89.8% precision and 93.3% NPV for ORM, and 81.7% precision and 85.1% NPV for PRM, substantially outperforming single VLMs and traditional ensemble approaches.

cs.SE↗

arXiv2025

Dipionic transitions of $Y(4500)$ to $J/ψ$

We, using an effective Lagrangian approach, investigated the dipionic transition of the $Y(4500)$ newly well established in the process $e^+e^-\to K^+K^- J/ψ$. In this study, the $Y(4500)$ was considered as a mixture of the $ψ(5S)$ and $ψ(4\,{^3D_1})$, and its dipionic decay to the $J/ψ$ was assumed to occur via charmed meson loops including box and two kinds of triangle loops. The calculated invariant mass spectra due to the different loop mechanisms exhibit distinct differences. These spectrum patterns are found to be nearly independent of the cutoff parameter, especially for the box and $\mathcal{R}$ loops. By means of comparing our calculated results with the future experimental measurements of the $π^+π^-$ and $π^\pm J/ψ$ invariant mass distributions, we could judge which loop mechanism is of more importance in this dipionic transition. The interference among the different kinds of loops is also exhibited. Despite the unknown phase angles between different kinds of loop diagrams, the partial decay width for the process $Y(4500)\toπ^+π^- J/ψ$ is estimated to be 250--700 keV, aligning with the estimation obtained by combining BESIII Collaboration measurements and theoretical predictions for $Y(4500)\to K^+K^- J/ψ$. We hope that the present calculations would be tested by the future BESIII or Belle II experiments.

hep-ph↗

arXiv2025

DAMSDAN: Distribution-Aware Multi-Source Domain Adaptation Network for Cross-Domain EEG-based Emotion Recognition

Significant inter-individual variability limits the generalization of EEG-based emotion recognition under cross-domain settings. We address two core challenges in multi-source adaptation: (1) dynamically modeling distributional heterogeneity across sources and quantifying their relevance to a target to reduce negative transfer; and (2) achieving fine-grained semantic consistency to strengthen class discrimination. We propose a distribution-aware multi-source domain adaptation network (DAMSDAN). DAMSDAN integrates prototype-based constraints with adversarial learning to drive the encoder toward discriminative, domain-invariant emotion representations. A domain-aware source weighting strategy based on maximum mean discrepancy (MMD) dynamically estimates inter-domain shifts and reweights source contributions. In addition, a prototype-guided conditional alignment module with dual pseudo-label interaction enhances pseudo-label reliability and enables category-level, fine-grained alignment, mitigating noise propagation and semantic drift. Experiments on SEED and SEED-IV show average accuracies of 94.86\% and 79.78\% for cross-subject, and 95.12\% and 83.15\% for cross-session protocols. On the large-scale FACED dataset, DAMSDAN achieves 82.88\% (cross-subject). Extensive ablations and interpretability analyses corroborate the effectiveness of the proposed framework for cross-domain EEG-based emotion recognition.

cs.LG↗

arXiv2025

Incentivizing Truthful Language Models via Peer Elicitation Games

Large Language Models (LLMs) have demonstrated strong generative capabilities but remain prone to inconsistencies and hallucinations. We introduce Peer Elicitation Games (PEG), a training-free, game-theoretic framework for aligning LLMs through a peer elicitation mechanism involving a generator and multiple discriminators instantiated from distinct base models. Discriminators interact in a peer evaluation setting, where utilities are computed using a determinant-based mutual information score that provably incentivizes truthful reporting without requiring ground-truth labels. We establish theoretical guarantees showing that each agent, via online learning, achieves sublinear regret in the sense their cumulative performance approaches that of the best fixed truthful strategy in hindsight. Moreover, we prove last-iterate convergence to a truthful Nash equilibrium, ensuring that the actual policies used by agents converge to stable and truthful behavior over time. Empirical evaluations across multiple benchmarks demonstrate significant improvements in factual accuracy. These results position PEG as a practical approach for eliciting truthful behavior from LLMs without supervision or fine-tuning.

cs.LG↗

arXiv2025

Uncertainty-Aware Extreme Point Tracing for Weakly Supervised Ultrasound Image Segmentation

Automatic medical image segmentation is a fundamental step in computer-aided diagnosis, yet fully supervised approaches demand extensive pixel-level annotations that are costly and time-consuming. To alleviate this burden, we propose a weakly supervised segmentation framework that leverages only four extreme points as annotation. Specifically, bounding boxes derived from the extreme points are used as prompts for the Segment Anything Model 2 (SAM2) to generate reliable initial pseudo labels. These pseudo labels are progressively refined by an enhanced Feature-Guided Extreme Point Masking (FGEPM) algorithm, which incorporates Monte Carlo dropout-based uncertainty estimation to construct a unified gradient uncertainty cost map for boundary tracing. Furthermore, a dual-branch Uncertainty-aware Scale Consistency (USC) loss and a box alignment loss are introduced to ensure spatial consistency and precise boundary alignment during training. Extensive experiments on two public ultrasound datasets, BUSI and UNS, demonstrate that our method achieves performance comparable to, and even surpassing fully supervised counterparts while significantly reducing annotation cost. These results validate the effectiveness and practicality of the proposed weakly supervised framework for ultrasound image segmentation.

cs.CV↗

arXiv2025

Learning from Disagreement: A Group Decision Simulation Framework for Robust Medical Image Segmentation

Medical image segmentation annotation suffers from inter-rater variability (IRV) due to differences in annotators' expertise and the inherent blurriness of medical images. Standard approaches that simply average expert labels are flawed, as they discard the valuable clinical uncertainty revealed in disagreements. We introduce a fundamentally new approach with our group decision simulation framework, which works by mimicking the collaborative decision-making process of a clinical panel. Under this framework, an Expert Signature Generator (ESG) learns to represent individual annotator styles in a unique latent space. A Simulated Consultation Module (SCM) then intelligently generates the final segmentation by sampling from this space. This method achieved state-of-the-art results on challenging CBCT and MRI datasets (92.11% and 90.72% Dice scores). By treating expert disagreement as a useful signal instead of noise, our work provides a clear path toward more robust and trustworthy AI systems for healthcare.

cs.CV↗

arXiv2025

Training-Free Group Relative Policy Optimization

Recent advances in Large Language Model (LLM) agents have demonstrated their promising general capabilities. However, their performance in specialized real-world domains often degrades due to challenges in effectively integrating external tools and specific prompting strategies. While methods like agentic reinforcement learning have been proposed to address this, they typically rely on costly parameter updates, for example, through a process that uses Supervised Fine-Tuning (SFT) followed by a Reinforcement Learning (RL) phase with Group Relative Policy Optimization (GRPO) to alter the output distribution. However, we argue that LLMs can achieve a similar effect on the output distribution by learning experiential knowledge as a token prior, which is a far more lightweight approach that not only addresses practical data scarcity but also avoids the common issue of overfitting. To this end, we propose Training-Free Group Relative Policy Optimization (Training-Free GRPO), a cost-effective solution that enhances LLM agent performance without any parameter updates. Our method leverages the group relative semantic advantage instead of numerical ones within each group of rollouts, iteratively distilling high-quality experiential knowledge during multi-epoch learning on a minimal ground-truth data. Such knowledge serves as the learned token prior, which is seamlessly integrated during LLM API calls to guide model behavior. Experiments on mathematical reasoning and web searching tasks demonstrate that Training-Free GRPO, when applied to DeepSeek-V3.1-Terminus, significantly improves out-of-domain performance. With just a few dozen training samples, Training-Free GRPO outperforms fine-tuned small LLMs with marginal training data and cost.

cs.CL↗

arXiv2025

Sub-GeV Right-Handed Neutrino as a Probe of Neutrino Mass Generation in the Minimal Left-Right Symmetric Model

The minimal left-right symmetric model (mLRSM) provides an elegant and testable framework for addressing the origin of neutrino masses. We examine the constraints on the sub-GeV right-handed (RH) neutrino in the type-II seesaw scenario of the mLRSM without left-right mixing, taking limits from collider searches, meson decays, supernovae, neutrinoless double beta ($0νββ$) decay and cosmology. Specifically, we derive the $0νββ$ decay constraints using the advanced effective field theory approach and up-to-date nuclear matrix element calculations. Besides, we update the SN1987A cooling bound with the state-of-the-art simulations, provide new constraints from the energy deposition in the supernova ejecta, and incorporate the stringent RH neutrino lifetime upper limit $τ\lesssim 0.023\text{ s}$ from the big bang nucleosynthesis. Our results identify the parameter region compatible with all current experimental and observational constraints, where the RH neutrino mass lies between 700 MeV and 1 GeV and the RH $W$ boson mass is slightly below 20 TeV. This region is exclusively probed by the future tonne-scale $0νββ$ decay experiments, providing a unique window to test the mLRSM and the possible origin of neutrino masses.

hep-ph↗

arXiv2025

Incentivizing Reasoning for Advanced Instruction-Following of Large Language Models

Existing large language models (LLMs) face challenges of following complex instructions, especially when multiple constraints are present and organized in paralleling, chaining, and branching structures. One intuitive solution, namely chain-of-thought (CoT), is expected to universally improve capabilities of LLMs. However, we find that the vanilla CoT exerts a negative impact on performance due to its superficial reasoning pattern of simply paraphrasing the instructions. It fails to peel back the compositions of constraints for identifying their relationship across hierarchies of types and dimensions. To this end, we propose RAIF, a systematic method to boost LLMs in dealing with complex instructions via incentivizing reasoning for test-time compute scaling. First, we stem from the decomposition of complex instructions under existing taxonomies and propose a reproducible data acquisition method. Second, we exploit reinforcement learning (RL) with verifiable rule-centric reward signals to cultivate reasoning specifically for instruction following. We address the shallow, non-essential nature of reasoning under complex instructions via sample-wise contrast for superior CoT enforcement. We also exploit behavior cloning of experts to facilitate steady distribution shift from fast-thinking LLMs to skillful reasoners. Extensive evaluations on seven comprehensive benchmarks confirm the validity of the proposed method, where a 1.5B LLM achieves 11.74% gains with performance comparable to a 8B LLM. Evaluation on OOD constraints also confirms the generalizability of our RAIF. Codes and data are available at https://github.com/yuleiqin/RAIF. Keywords: reinforcement learning with verifiable rewards (RLVR), instruction following, complex instructions

cs.CV↗

arXiv2025

RoRecomp: Enhancing Reasoning Efficiency via Rollout Response Recomposition in Reinforcement Learning

Reinforcement learning with verifiable rewards (RLVR) has proven effective in eliciting complex reasoning in large language models (LLMs). However, standard RLVR training often leads to excessively verbose processes (in reasoning tasks) and inefficient exploration trajectories (in agentic settings), as outcome-only rewards provide no incentive for efficiency and the high variance in response length within relatively small rollout groups results in noisy optimization signals. To address this, we propose Rollout Response Recomposition (RoRecomp), a plug-and-play method that guides models toward concise reasoning by strategically recomposing the training data. RoRecomp separates responses into two distinct batch types: 1) priority batches, which combine short-correct and long-incorrect responses selected from online batches to provide a clear gradient signal for brevity, and 2) compensation batches, which utilize remaining responses from a replay buffer to maintain stability and prevent model collapse. To comprehensively evaluate effectiveness, we test RoRecomp across three settings where results demonstrate substantial efficiency gains: reducing reasoning length by 27.7% in zero RL training, reducing unnecessary tool calls by 46.8% while improving accuracy in agentic RL, and achieving up to 52.5% length reduction in thinking compression, all with minimal performance impact.

cs.AI↗

arXiv2025

Digital Privacy Under Attack: Challenges and Enablers

We present a comprehensive analysis of privacy attacks and countermeasures in data-driven systems. We systematically categorize attacks targeting three domains: anonymous data (linkage and structural attacks), statistical aggregates (reconstruction and differential attacks), and privacy-preserving models (extraction, reconstruction, membership inference, and inversion attacks). For each category, we analyze attack methodologies, adversary capabilities, and vulnerability mechanisms. We further evaluate countermeasures including perturbation techniques, randomization methods, query auditing, and model-level defenses, examining their effectiveness and inherent privacy-utility tradeoffs. Our analysis reveals that while differential privacy offers strong theoretical guarantees, it faces implementation challenges and potential vulnerabilities to emerging attacks. We identify critical research directions and provide researchers and practitioners with a structured framework for understanding privacy resilience in increasingly complex data ecosystems.

cs.CR↗

arXiv2025

Hadronic decays of possible pseudoscalar $P$-wave $D\bar{D}^\ast/\bar{D}D^\ast$ molecular state

Recently, the new structure $G(3900)$ observed by the BESIII Collaboration in the $e^+e^-\to D\bar{D}$ was identified to be the $P$-wave $D\bar{D}^\ast/\bar{D}D^\ast$ vector molecular resonance using a unified meson exchange model. Apart from the vector $P$-wave state, a possible pseudoscalar $P$-wave molecular state of the $D\bar{D}^\ast/\bar{D}D^\ast$ [called $G_0(3900)$ for short] was also predicted, which is likely to be observed in future experiments. Within the molecular framework, we calculated the partial decay widths for a series of hadronic decays of the $G_0$, including $G_0\to ω(ρ^0) J/ψ$, $π^+π^- η_c (1S)$, $π^+π^- χ_{c1} (1P)$, and $D^0\bar{D}^0π^0$. Under present model parameters, the hidden-charm decay modes are dominated by the $G_0\toωJ/ψ$ and $G_0\toπ^+π^-η_c (1S)$, and the partial widths can reach 1 MeV and 0.1 MeV, respectively. The open-charm channel $G_0\to D^0\bar{D}^0π^0$ exhibits a rather small decay rate ($\sim 0.1~\mathrm{keV}$). In terms of our present predictions, we suggest BESIII and Belle II to search for the pseudoscalar $P$-wave $D\bar{D}^\ast / \bar{D}D^\ast$ molecular state with $J^{PC} = 0^{-+}$ in the hidden-charm processes $G_0 \to ωJ/ψ$ or $G_0 \to π^+π^-η_c(1S)$.

hep-ph↗

arXiv2025

Memory in Large Language Models: Mechanisms, Evaluation and Evolution

Under a unified operational definition, we define LLM memory as a persistent state written during pretraining, finetuning, or inference that can later be addressed and that stably influences outputs. We propose a four-part taxonomy (parametric, contextual, external, procedural/episodic) and a memory quadruple (location, persistence, write/access path, controllability). We link mechanism, evaluation, and governance via the chain write -> read -> inhibit/update. To avoid distorted comparisons across heterogeneous setups, we adopt a three-setting protocol (parametric only, offline retrieval, online retrieval) that decouples capability from information availability on the same data and timeline. On this basis we build a layered evaluation: parametric (closed-book recall, edit differential, memorization/privacy), contextual (position curves and the mid-sequence drop), external (answer correctness vs snippet attribution/faithfulness), and procedural/episodic (cross-session consistency and timeline replay, E MARS+). The framework integrates temporal governance and leakage auditing (freshness hits, outdated answers, refusal slices) and uncertainty reporting via inter-rater agreement plus paired tests with multiple-comparison correction. For updating and forgetting, we present DMM Gov: coordinating DAPT/TAPT, PEFT, model editing (ROME, MEND, MEMIT, SERAC), and RAG to form an auditable loop covering admission thresholds, rollout, monitoring, rollback, and change audits, with specs for timeliness, conflict handling, and long-horizon consistency. Finally, we give four testable propositions: minimum identifiability; a minimal evaluation card; causally constrained editing with verifiable forgetting; and when retrieval with small-window replay outperforms ultra-long-context reading. This yields a reproducible, comparable, and governable coordinate system for research and deployment.

cs.AI↗

arXiv2025

When marine radar target detection meets pretrained large language models

Deep learning (DL) methods are widely used to extract high-dimensional patterns from the sequence features of radar echo signals. However, conventional DL algorithms face challenges such as redundant feature segments, and constraints from restricted model sizes. To address these issues, we propose a framework that integrates feature preprocessing with large language models (LLMs). Our preprocessing module tokenizes radar sequence features, applies a patch selection algorithm to filter out uninformative segments, and projects the selected patches into embeddings compatible with the feature space of pre-trained LLMs. Leveraging these refined embeddings, we incorporate a pre-trained LLM, fine-tuning only the normalization layers to reduce training burdens while enhancing performance. Experiments on measured datasets demonstrate that the proposed method significantly outperforms the state-of-the-art baselines on supervised learning tests.

eess.SP↗