arXiv subjects

Bin Hu

Publications and source records attributed to Bin Hu.

At least 109 records · Page 6Linked to original sources

arXiv2025

Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model

We present Ring-1T, the first open-source, state-of-the-art thinking model with a trillion-scale parameter. It features 1 trillion total parameters and activates approximately 50 billion per token. Training such models at a trillion-parameter scale introduces unprecedented challenges, including train-inference misalignment, inefficiencies in rollout processing, and bottlenecks in the RL system. To address these, we pioneer three interconnected innovations: (1) IcePop stabilizes RL training via token-level discrepancy masking and clipping, resolving instability from training-inference mismatches; (2) C3PO++ improves resource utilization for long rollouts under a token budget by dynamically partitioning them, thereby obtaining high time efficiency; and (3) ASystem, a high-performance RL framework designed to overcome the systemic bottlenecks that impede trillion-parameter model training. Ring-1T delivers breakthrough results across critical benchmarks: 93.4 on AIME-2025, 86.72 on HMMT-2025, 2088 on CodeForces, and 55.94 on ARC-AGI-1. Notably, it attains a silver medal-level result on the IMO-2025, underscoring its exceptional reasoning capabilities. By releasing the complete 1T parameter MoE model to the community, we provide the research community with direct access to cutting-edge reasoning capabilities. This contribution marks a significant milestone in democratizing large-scale reasoning intelligence and establishes a new baseline for open-source model performance.

cs.CL↗

arXiv2025

Beyond Seeing: Evaluating Multimodal LLMs on Tool-Enabled Image Perception, Transformation, and Reasoning

Multimodal Large Language Models (MLLMs) are increasingly applied in real-world scenarios where user-provided images are often imperfect, requiring active image manipulations such as cropping, editing, or enhancement to uncover salient visual cues. Beyond static visual perception, MLLMs must also think with images: dynamically transforming visual content and integrating it with other tools to solve complex tasks. However, this shift from treating vision as passive context to a manipulable cognitive workspace remains underexplored. Most existing benchmarks still follow a think about images paradigm, where images are regarded as static inputs. To address this gap, we introduce VisualToolBench, a visual tool-use reasoning benchmark that rigorously evaluates MLLMs' ability to perceive, transform, and reason across complex visual-textual tasks under the think-with-images paradigm. VisualToolBench comprises 1,204 challenging, open-ended vision tasks (603 single-turn, 601 multi-turn) spanning across five diverse domains, each paired with detailed rubrics to enable systematic evaluation. Our evaluation shows that current MLLMs struggle with tasks requiring effective integration of vision and general-purpose tools. Even the strongest model (GPT-5-think) reaches only 18.68% pass rate. We further observe divergent tool-use behaviors, with OpenAI models benefiting from diverse image manipulations while Gemini-2.5-pro shows no improvement. By introducing the first benchmark centered on think with images, VisualToolBench offers critical insights for advancing visual intelligence in MLLMs.

cs.CV↗

arXiv2025

From Generic to Specialized: A Subspecialty Diagnostic System Powered by Self-Supervised Learning for Cervical Histopathology

Cervical cancer remains a major malignancy, necessitating extensive and complex histopathological assessments and comprehensive support tools. Although deep learning shows promise, these models still lack accuracy and generalizability. General foundation models offer a broader reach but remain limited in capturing subspecialty-specific features and task adaptability. We introduce the Cervical Subspecialty Pathology (CerS-Path) diagnostic system, developed through two synergistic pretraining stages: self-supervised learning on approximately 190 million tissue patches from 140,000 slides to build a cervical-specific feature extractor, and multimodal enhancement with 2.5 million image-text pairs, followed by integration with multiple downstream diagnostic functions. Supporting eight diagnostic functions, including rare cancer classification and multimodal Q&A, CerS-Path surpasses prior foundation models in scope and clinical applicability. Comprehensive evaluations demonstrate a significant advance in cervical pathology, with prospective testing on 3,173 cases across five centers maintaining 99.38% screening sensitivity and excellent generalizability, highlighting its potential for subspecialty diagnostic translation and cervical cancer screening.

cs.CV↗

arXiv2025

Artificial intelligence as a surrogate brain: Bridging neural dynamical models and data

Recent breakthroughs in artificial intelligence (AI) are reshaping the way we construct computational counterparts of the brain, giving rise to a new class of ``surrogate brains''. In contrast to conventional hypothesis-driven biophysical models, the AI-based surrogate brain encompasses a broad spectrum of data-driven approaches to solve the inverse problem, with the primary objective of accurately predicting future whole-brain dynamics with historical data. Here, we introduce a unified framework of constructing an AI-based surrogate brain that integrates forward modeling, inverse problem solving, and model evaluation. Leveraging the expressive power of AI models and large-scale brain data, surrogate brains open a new window for decoding neural systems and forecasting complex dynamics with high dimensionality, nonlinearity, and adaptability. We highlight that the learned surrogate brain serves as a simulation platform for dynamical systems analysis, virtual perturbation, and model-guided neurostimulation. We envision that the AI-based surrogate brain will provide a functional bridge between theoretical neuroscience and translational neuroengineering.

q-bio.NC↗

arXiv2025

Accelerating the Monte Carlo simulation of the Enskog equation for multiscale dense gas flows

A general synthetic iterative scheme is proposed to solve the Enskog equation within a Monte Carlo framework. The method demonstrates rapid convergence by reducing intermediate Monte Carlo evolution and preserves the asymptotic-preserving property, enabling spatial cell sizes much larger than the mean free path in near-continuum flows. This is realized through mesoscopic-macroscopic two-way coupling: the mesoscopic Monte Carlo simulation provides high-order constitutive relations to close the moment (synthetic) equation, while the macroscopic synthetic equation, once solved toward steady state, directs the evolution of simulation particles in the Monte Carlo method. The accuracy of the proposed general synthetic iterative scheme is verified through one-dimensional normal shock wave and planar Fourier heat transfer problems, while its fast-converging and asymptotic-preserving properties are demonstrated in the force-driven Poiseuille flow and two-dimensional hypersonic cylinder flow and low-speed porous media flow, where the simulation time is reduced by several orders of magnitude in near-continuum flows. With the proposed method, a brief analysis is conducted on the role of the adsorption layer in porous media flow, mimicking shale gas extraction.

physics.flu-dyn↗

arXiv2025

Detectability of massive binary black holes with sub-mHz gravitational wave missions

Beyond LISA, proposed space-based gravitational wave (GW) missions aim to explore the sub-millihertz to microhertz frequency band, with one key objective being the detection of massive binary black hole (MBBH) mergers across cosmic distances. In this work, we investigate the detection and localization capabilities of future sub-mHz GW observatories for MBBH coalescences. Including the full galactic foreground noise, we find that signal-to-noise ratios (SNRs) can reach several thousand across a wide range of redshifts. We evaluate three representative orbital configurations--non-precessing and precessing with different inclination angles--and analyze their localization performance for various MBBH populations. In the non-precessing case, a two-hemisphere degeneracy arises when only the dominant (2,2) mode is considered, which is effectively resolved by including higher-order modes. These modes contribute to a more uniform performance across all configurations, thereby mitigating the prior advantage of precessing mission orbits. Sub-mHz missions operating in the [10 $μ$Hz, 10 mHz] band partially overlap with LISA's range but provide enhanced sensitivity to lower-frequency GWs due to their longer interferometric baselines. This results in significantly improved localization of high-mass MBBHs, enhancing the prospects for multi-messenger astronomy and precision cosmology. Moreover, the high SNRs attainable with sub-mHz detectors could enable stringent tests of general relativity and alternative theories of gravity.

gr-qc↗

arXiv2025

SOLAR: Switchable Output Layer for Accuracy and Robustness in Once-for-All Training

Once-for-All (OFA) training enables a single super-net to generate multiple sub-nets tailored to diverse deployment scenarios, supporting flexible trade-offs among accuracy, robustness, and model-size without retraining. However, as the number of supported sub-nets increases, excessive parameter sharing in the backbone limits representational capacity, leading to degraded calibration and reduced overall performance. To address this, we propose SOLAR (Switchable Output Layer for Accuracy and Robustness in Once-for-All Training), a simple yet effective technique that assigns each sub-net a separate classification head. By decoupling the logit learning process across sub-nets, the Switchable Output Layer (SOL) reduces representational interference and improves optimization, without altering the shared backbone. We evaluate SOLAR on five datasets (SVHN, CIFAR-10, STL-10, CIFAR-100, and TinyImageNet) using four super-net backbones (ResNet-34, WideResNet-16-8, WideResNet-40-2, and MobileNetV2) for two OFA training frameworks (OATS and SNNs). Experiments show that SOLAR outperforms the baseline methods: compared to OATS, it improves accuracy of sub-nets up to 1.26 %, 4.71 %, 1.67 %, and 1.76 %, and robustness up to 9.01 %, 7.71 %, 2.72 %, and 1.26 % on SVHN, CIFAR-10, STL-10, and CIFAR-100, respectively. Compared to SNNs, it improves TinyImageNet accuracy by up to 2.93 %, 2.34 %, and 1.35 % using ResNet-34, WideResNet-16-8, and MobileNetV2 backbones (with 8 sub-nets), respectively.

cs.LG↗

arXiv2025

Introduction to the Chinese Space Station Survey Telescope (CSST)

The Chinese Space Station Survey Telescope (CSST) is an upcoming Stage-IV sky survey telescope, distinguished by its large field of view (FoV), high image quality, and multi-band observation capabilities. It can simultaneously conduct precise measurements of the Universe by performing multi-color photometric imaging and slitless spectroscopic surveys. The CSST is equipped with five scientific instruments, i.e. Multi-band Imaging and Slitless Spectroscopy Survey Camera (SC), Multi-Channel Imager (MCI), Integral Field Spectrograph (IFS), Cool Planet Imaging Coronagraph (CPI-C), and THz Spectrometer (TS). Using these instruments, CSST is expected to make significant contributions and discoveries across various astronomical fields, including cosmology, galaxies and active galactic nuclei (AGN), the Milky Way and nearby galaxies, stars, exoplanets, Solar System objects, astrometry, and transients and variable sources. This review aims to provide a comprehensive overview of the CSST instruments, observational capabilities, data products, and scientific potential.

astro-ph.IM↗

arXiv2025

When Fine-Tuning is Not Enough: Lessons from HSAD on Hybrid and Adversarial Audio Spoof Detection

The rapid advancement of AI has enabled highly realistic speech synthesis and voice cloning, posing serious risks to voice authentication, smart assistants, and telecom security. While most prior work frames spoof detection as a binary task, real-world attacks often involve hybrid utterances that mix genuine and synthetic speech, making detection substantially more challenging. To address this gap, we introduce the Hybrid Spoofed Audio Dataset (HSAD), a benchmark containing 1,248 clean and 41,044 degraded utterances across four classes: human, cloned, zero-shot AI-generated, and hybrid audio. Each sample is annotated with spoofing method, speaker identity, and degradation metadata to enable fine-grained analysis. We evaluate six transformer-based models, including spectrogram encoders (MIT-AST, MattyB95-AST) and self-supervised waveform models (Wav2Vec2, HuBERT). Results reveal critical lessons: pretrained models overgeneralize and collapse under hybrid conditions; spoof-specific fine-tuning improves separability but struggles with unseen compositions; and dataset-specific adaptation on HSAD yields large performance gains (AST greater than 97 percent and F1 score is approximately 99 percent), though residual errors persist for complex hybrids. These findings demonstrate that fine-tuning alone is not sufficient-robust hybrid-aware benchmarks like HSAD are essential to expose calibration failures, model biases, and factors affecting spoof detection in adversarial environments. HSAD thus provides both a dataset and an analytic framework for building resilient and trustworthy voice authentication systems.

cs.SD↗

arXiv2025

VLASCD: A Visual Language Action Model for Simultaneous Chatting and Decision Making

Recent large pretrained models such as LLMs (e.g., GPT series) and VLAs (e.g., OpenVLA) have achieved notable progress on multimodal tasks, yet they are built upon a multi-input single-output (MISO) paradigm. We show that this paradigm fundamentally limits performance in multi-input multi-output (MIMO) scenarios, where parallel task execution is required. In MISO architectures, tasks compete for a shared output channel, creating mutual exclusion effects that cause unbalanced optimization and degraded performance. To address this gap, we introduce MIMO-VLA (VLASCD), a unified training framework that enables concurrent multi-task outputs, exemplified by simultaneous dialogue generation and decision-making. Inspired by human cognition, MIMO-VLA eliminates interference between tasks and supports efficient parallel processing. Experiments on the CARLA autonomous driving platform demonstrate that MIMO-VLA substantially outperforms state-of-the-art MISO-based LLMs, reinforcement learning models, and VLAs in MIMO settings, establishing a new direction for multimodal and multitask learning.

cs.AI↗

arXiv2025

Large Foundation Model for Ads Recommendation

Online advertising relies on accurate recommendation models, with recent advances using pre-trained large-scale foundation models (LFMs) to capture users' general interests across multiple scenarios and tasks. However, existing methods have critical limitations: they extract and transfer only user representations (URs), ignoring valuable item representations (IRs) and user-item cross representations (CRs); and they simply use a UR as a feature in downstream applications, which fails to bridge upstream-downstream gaps and overlooks more transfer granularities. In this paper, we propose LFM4Ads, an All-Representation Multi-Granularity transfer framework for ads recommendation. It first comprehensively transfers URs, IRs, and CRs, i.e., all available representations in the pre-trained foundation model. To effectively utilize the CRs, it identifies the optimal extraction layer and aggregates them into transferable coarse-grained forms. Furthermore, we enhance the transferability via multi-granularity mechanisms: non-linear adapters for feature-level transfer, an Isomorphic Interaction Module for module-level transfer, and Standalone Retrieval for model-level transfer. LFM4Ads has been successfully deployed in Tencent's industrial-scale advertising platform, processing tens of billions of daily samples while maintaining terabyte-scale model parameters with billions of sparse embedding keys across approximately two thousand features. Since its production deployment in Q4 2024, LFM4Ads has achieved 10+ successful production launches across various advertising scenarios, including primary ones like Weixin Moments and Channels. These launches achieve an overall GMV lift of 2.45% across the entire platform, translating to estimated annual revenue increases in the hundreds of millions of dollars.

cs.LG↗

arXiv2025

Developing an Inhaled NEU1 Inhibitor for Cystic Fibrosis via Pharmacokinetic and Biophysical Modeling

Background: Cystic fibrosis (CF) airway mucus exhibits reduced mucin sialylation, increasing viscosity and impairing mucociliary clearance (MCC). NEU1 inhibition has been proposed to restore MCC, but its quantitative pharmacokinetic and rheological effects, particularly with inhaled delivery, remain uncharacterized. Objective: To develop an integrated pharmacokinetic/pharmacodynamic (PK/PD) and biophysical model to assess the efficacy of an inhaled NEU1 inhibitor. Methods: Empirical and preclinical NEU1 inhibition data were combined with inhalation PK/PD modeling and a biophysical viscosity framework linking mucin sialylation and extracellular DNA. Synthetic cohort simulations (N = 200) were reconciled with empirical PK benchmarks using Latin hypercube parameter sampling. Cross-validation, hold-out testing, and causal inference methods (inverse probability of treatment weighting and targeted maximum likelihood estimation) quantified predicted effects on lung function (delta FEV1). Results: With reconciled parameters (F_dep = 0.12; k_abs = 0.21 per hour; k_muc = 0.24 per hour), epithelial lining fluid drug levels reached a peak concentration of 7.5 micromolar (95 percent CI: 6 to 10 micromolar), achieving IC50 coverage for approximately 10 hours per day and greater than 80 percent modeled NEU1 inhibition. Predicted mucus viscosity reduction averaged 25 to 28 percent. Causal inference estimated delta FEV1 improvement of +0.13 liters (95 percent CI: 0.10 to 0.15 liters), with about 70 percent mediated via MCC. Conclusions: Empirically anchored PK/PD and biophysical modeling support the feasibility of inhaled NEU1 inhibition as a rheology-targeting strategy in CF, projecting clinically realistic efficacy while maintaining pharmacological viability. This calibrated proof of concept warrants in vivo validation in CF models.

q-bio.QM↗

arXiv2025

Linking GFAP Levels to Speech Anomalies in Acute Brain Injury: A Simulation Based Study

Background: Glial fibrillary acidic protein (GFAP) is a biomarker for intracerebral hemorrhage and traumatic brain injury, but its link to acute speech disruption is untested. Speech anomalies often emerge early after injury, enabling rapid triage. Methods: We simulated a cohort of 200 virtual patients stratified by lesion location, onset time, and severity. GFAP kinetics followed published trajectories; speech anomalies were generated from lesion-specific neurophysiological mappings. Ensemble machine-learning models used GFAP, speech, and lesion features; robustness was tested under noise, delays, and label dropout. Causal inference (inverse probability of treatment weighting and targeted maximum likelihood estimation) estimated directional associations between GFAP elevation and speech severity. Findings: GFAP correlated with simulated speech anomaly severity (Spearman rho = 0.48), strongest for cortical lesions (rho = 0.55). Voice anomalies preceded detectable GFAP rise by a median of 42 minutes in cortical injury. Classifier area under the curve values were 0.74 (GFAP only), 0.78 (voice only), and 0.86 for the fused multimodal model, which showed higher sensitivity in mild or ambiguous cases. Causal estimates indicated higher GFAP increased the modeled probability of moderate-to-severe speech anomalies by 32 to 35 percent, independent of lesion site and onset time. Conclusion: These results support a link between GFAP elevation and speech anomalies in acute brain injury and suggest integrated biochemical-voice diagnostics could improve early triage, especially for cortical injury. Findings are simulation-based and require validation in prospective clinical studies with synchronized GFAP assays and speech recordings.

q-bio.NC↗

arXiv2025

CM-Diff: A Single Generative Network for Bidirectional Cross-Modality Translation Diffusion Model Between Infrared and Visible Images

Image translation is one of the crucial approaches for mitigating information deficiencies in the infrared and visible modalities, while also facilitating the enhancement of modality-specific datasets. However, existing methods for infrared and visible image translation either achieve unidirectional modality translation or rely on cycle consistency for bidirectional modality translation, which may result in suboptimal performance. In this work, we present the bidirectional cross-modality translation diffusion model (CM-Diff) for simultaneously modeling data distributions in both the infrared and visible modalities. We address this challenge by combining translation direction labels for guidance during training with cross-modality feature control. Specifically, we view the establishment of the mapping relationship between the two modalities as the process of learning data distributions and understanding modality differences, achieved through a novel Bidirectional Diffusion Training (BDT). Additionally, we propose a Statistical Constraint Inference (SCI) to ensure the generated image closely adheres to the data distribution of the target modality. Experimental results demonstrate the superiority of our CM-Diff over state-of-the-art methods, highlighting its potential for generating dual-modality datasets.

cs.CV↗

arXiv2025

How LLMs Comprehend Temporal Meaning in Narratives: A Case Study in Cognitive Evaluation of LLMs

Large language models (LLMs) exhibit increasingly sophisticated linguistic capabilities, yet the extent to which these behaviors reflect human-like cognition versus advanced pattern recognition remains an open question. In this study, we investigate how LLMs process the temporal meaning of linguistic aspect in narratives that were previously used in human studies. Using an Expert-in-the-Loop probing pipeline, we conduct a series of targeted experiments to assess whether LLMs construct semantic representations and pragmatic inferences in a human-like manner. Our findings show that LLMs over-rely on prototypicality, produce inconsistent aspectual judgments, and struggle with causal reasoning derived from aspect, raising concerns about their ability to fully comprehend narratives. These results suggest that LLMs process aspect fundamentally differently from humans and lack robust narrative understanding. Beyond these empirical findings, we develop a standardized experimental framework for the reliable assessment of LLMs' cognitive and linguistic capabilities.

cs.CL↗

arXiv2025

Residual test to search for microlensing signatures in strongly lensed gravitational wave signals

When a gravitational wave signal encounters a massive object, such as a galaxy or galaxy cluster, it undergoes strong gravitational lensing, producing multiple copies of the original signal. These strongly lensed signals exhibit identical waveform morphology in the frequency domain, allowing analysis without the need for complex lens models. However, stellar fields and dark matter substructures within the galactic lens introduce microlensing effects that alter individual signal morphologies. Identifying these microlensing signatures is computationally challenging within Bayesian frameworks. In this study, we propose a residual test to efficiently search for microlensing signatures by leveraging the fact that current Bayesian inference pipelines are optimized solely for the strong lensing hypothesis. Using cross-correlation techniques, we investigate the microlensing-induced deviations from the strong hypothesis, which are imprinted in the residuals. Most simulated signals from our realistic microlensing populations exhibit small mismatches between the microlensed and unlensed waveforms, but a fraction show significant deviations. We find that 28% (52%) and 34% (66%)of microlensed events with mismatch > 0.03 and > 0.1, respectively, can be discerned with O4 (O5) detector sensitivities, which demonstrates that high-mismatch events are more likely to be identified as microlensed. Including all events from a realistic population, 11% (21.5%) are identifiable with O4 (O5) sensitivity using our approach.

gr-qc↗

arXiv2025

Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMs

We present Ring-lite, a Mixture-of-Experts (MoE)-based large language model optimized via reinforcement learning (RL) to achieve efficient and robust reasoning capabilities. Built upon the publicly available Ling-lite model, a 16.8 billion parameter model with 2.75 billion activated parameters, our approach matches the performance of state-of-the-art (SOTA) small-scale reasoning models on challenging benchmarks (e.g., AIME, LiveCodeBench, GPQA-Diamond) while activating only one-third of the parameters required by comparable models. To accomplish this, we introduce a joint training pipeline integrating distillation with RL, revealing undocumented challenges in MoE RL training. First, we identify optimization instability during RL training, and we propose Constrained Contextual Computation Policy Optimization(C3PO), a novel approach that enhances training stability and improves computational throughput via algorithm-system co-design methodology. Second, we empirically demonstrate that selecting distillation checkpoints based on entropy loss for RL training, rather than validation metrics, yields superior performance-efficiency trade-offs in subsequent RL training. Finally, we develop a two-stage training paradigm to harmonize multi-domain data integration, addressing domain conflicts that arise in training with mixed dataset. We will release the model, dataset, and code.

cs.CL↗

arXiv2025

A brain-inspired generative model for EEG-based cognitive state identification

This article proposes a brain-inspired generative (BIG) model that merges an impulsive-attention neural network and a variational autoencoder (VAE) for identifying cognitive states based on electroencephalography (EEG) data. A hybrid learning method is presented for training the model by integrating gradient-based learning and heteroassociative memory. The BIG model is capable of achieving multi-task objectives: EEG classification, generating new EEG, and brain network interpretation, alleviating the limitations of excessive data training and high computational cost in conventional approaches. Experimental results on two public EEG datasets with different sampling rates demonstrate that the BIG model achieves a classification accuracy above 89\%, comparable with state-of-the-art methods, while reducing computational cost by nearly 11\% over the baseline EEGNet. Incorporating the generated EEG data for training, the BIG model exhibits comparative performance in a few-shot pattern. Ablation studies justify the poised brain-inspired characteristic regarding the impulsive-attention module and the hybrid learning method. Thanks to the performance advantages with interpretable outputs, this BIG model has application potential for building digital twins of the brain.

eess.SP↗