arXiv subjects

Chen Hu

Publications and source records attributed to Chen Hu.

At least 19 recordsLinked to original sources

arXiv2026

Automatic Model-Hardware Co-Adaptation for Heterogeneous AI Accelerators

Large language models now evolve faster than production inference systems can be ported and optimized. New releases change attention, MoE routing, quantization formats, KV-cache layout, and parallel execution patterns, while deployed accelerator fleets remain heterogeneous across hardware generations, framework forks, operator libraries, compiler backends, and communication runtimes. Serving a new model on existing hardware is therefore a model-framework-kernel-hardware co-adaptation problem. We present MetaInfer, an LLM-agent system that formulates inference adaptation as route search over a costed execution-adaptation graph. The graph connects model semantics, framework dispatch, kernel choices, hardware capabilities, runtime evidence, and serving objectives. MetaInfer constructs and updates this graph during execution, restores missing or blocked routes through patches, and reduces route cost through staged validation and end-to-end profiling. Three real episodes -- DeepSeek V4 Flash on NVIDIA A800, GLM 5.3 Flash on NVIDIA A800, and DeepSeek V4 Flash on Hygon K100AI DCU -- demonstrate deployment repair, cross-model knowledge transfer, and portability across heterogeneous accelerator software stacks.

cs.MA↗

arXiv2026

Rethinking Scientific Discovery in the Agentic Era

Artificial intelligence has advanced scientific discovery, but most AI4Science systems remain fragmented tools that rely on humans to coordinate problem formulation, literature grounding, model use, simulation, validation, and knowledge reuse. This paper presents \textbf{SCION (Scientific Collaborative Innovation with Agentic Organizational Nexus)}, an agentic scientific operating system that acts as an \textbf{organizational nexus}. Through a Science Agent serving as a \textbf{Meta-Harness}, SCION connects scientific tasks, tools, agents, artifacts, and memory, transforming research into an executable, auditable, and reusable operational process. At its core is the \textbf{Research Execution Plan (REP)}, which compiles high-level scientific intent into staged objectives, dependencies, verification checkpoints, tool requirements, expected artifacts, and fallback conditions. SCION further integrates hierarchical multi-agent execution, profile-driven specialization, selective context construction, governed delegation, and layered epistemic memory to support long-horizon scientific work. We formulate discovery under SCION as \textbf{Target-conditioned Inverse Search} and extend it to hidden-target settings through batch active search under finite experimental budgets. Applications in materials analysis, molecule design, and protein or antibody screening, together with experiments on scientific reading, idea generation, molecule generation, and antibody screening, show that SCION outperforms existing autonomous research-agent baselines, especially in decomposition, verification, refinement, and memory reuse. Overall, SCION shifts AI from isolated tools toward a coordinated operational layer for traceable and reusable scientific innovation.

cs.CL↗

arXiv2026

Mrk 382: A Narrow-line Seyfert 1 Galaxy with Recurrent X-ray State Transitions

We report recurrent X-ray state transitions in the nearby narrow-line Seyfert~1 galaxy Mrk~382 using multi-epoch observations from \textit{Swift}, \textit{Chandra}, \textit{XMM-Newton}, and eROSITA, together with archival ultraviolet, optical, and infrared data. The 0.3--2 keV flux varies by nearly an order of magnitude over the past $\sim15$ yr, with multiple transitions between bright and faint states. The source brightened by a factor of $\sim10$ between the 2010 \textit{Chandra} observation and the 2011 \textit{XMM-Newton} high state, then declined by $\sim6$--7 to a low state in 2019, followed by renewed brightening in recent \textit{Swift} monitoring. The X-ray spectrum shows strong state-dependent evolution, changing from a steep high-state continuum ($\Gamma=2.32\pm0.04$) to a much harder low-state spectrum ($\Gamma=1.39\pm0.06$). The low-state spectrum also exhibits a narrow Fe K$\alpha$ line with an equivalent width of $\sim330$ eV. Reflection modeling indicates that the low-flux state is strongly reflection dominated, with the reflection fraction increasing from $R_{\rm refl}\sim4$ to $\sim34$, consistent with a compact corona subject to strong light-bending effects. The ultraviolet emission broadly follows the long-term X-ray variability but with smaller amplitude, while the optical and mid-infrared bands vary more mildly. Despite the dramatic X-ray variability, Mrk~382 does not enter an extreme X-ray-weak state, and we did not detect clear optical spectral-type changes based on the currently available observations. Mrk~382 is therefore a rare nearby Seyfert galaxy undergoing recurrent X-ray state transitions, providing a valuable laboratory for studying changing coronal geometry and multiwavelength AGN variability.

astro-ph.HE↗

arXiv2026

A Sliced-Wasserstein Framework on Correlation Matrices for EEG Decoding

Electroencephalography (EEG) offers noninvasive, millisecond resolution recordings of neuronal activity and is widely used in neuroscience and healthcare. Many EEG decoding pipelines rely on covariance descriptors for their robustness to noise, but such representations are sensitive to channel-wise scaling. Recent studies have therefore advocated full-rank correlation matrices as a scale-invariant alternative for EEG decoding. In this paper, we study Sliced-Wasserstein (SW) discrepancies for probability distributions on the manifold of full-rank correlation matrices. We adopt the pullback-Euclidean formulation of SW, referred to as Pullback Euclidean Metric Sliced-Wasserstein (PEMSW), and instantiate it under two recently introduced correlation geometries, \textit{i.e.}, the Off-Log Metric (OLM) and Log-Scaled Metric (LSM). This yields two Correlation Sliced-Wasserstein (CorSW) discrepancies with closed-form slicing coordinates and efficient computation through one-dimensional Wasserstein distances. Building on CorSW, we further develop a domain generalization (DG) framework for EEG decoding. Experiments on three EEG datasets demonstrate improved generalization under distribution shifts, with low training overhead and no additional inference cost. The source code is available at github.com/ChenHu-ML/CorSW.

cs.LG↗

arXiv2026

CoPA: Benchmarking Personalized Question Answering with Data-Informed Cognitive Factors

While LLMs have demonstrated remarkable potential in Question Answering (QA), evaluating personalization remains a critical bottleneck. Existing paradigms predominantly rely on lexical-level similarity or manual heuristics, often lacking sufficient data-driven validation. We address this by mining Community-Individual Preference Divergence (CIPD), where individual choices override consensus, to distill six key personalization factors as evaluative dimensions. Accordingly, we introduce CoPA, a benchmark with 1,985 user profiles for fine-grained, factor-level assessment. By quantifying the alignment between model outputs and user-specific cognitive preferences inferred from interaction patterns, CoPA provides a more comprehensive and discriminative standard for evaluating personalized QA than generic metrics. The code is available at https://github.com/bjzgcai/CoPA.

cs.CL↗

arXiv2026

MIXAR: Scaling Autoregressive Pixel-based Language Models to Multiple Languages and Scripts

Pixel-based language models are gaining momentum as alternatives to traditional token-based approaches, promising to circumvent tokenization challenges. However, the inherent perceptual diversity across languages poses a significant hurdle for multilingual generalization in pixel space. This paper introduces MIXAR, the first generative pixel-based language model trained on eight different languages utilizing a range of different scripts. We empirically evaluate MIXAR against previous pixel-based models as well as comparable tokenizer-based models, demonstrating substantial performance improvement on discriminative and generative multilingual tasks. Additionally, we show how MIXAR is robust to languages never seen during the training. These results are further strengthened when scaling the model to 0.5B parameters which not only improves its capabilities in generative tasks like LAMBADA but also its robustness when challenged with input perturbations such as orthographic attacks.

cs.CL↗

arXiv2026

Radiology Report Generation for Low-Quality X-Ray Images

Vision-Language Models (VLMs) have significantly advanced automated Radiology Report Generation (RRG). However, existing methods implicitly assume high-quality inputs, overlooking the noise and artifacts prevalent in real-world clinical environments. Consequently, current models exhibit severe performance degradation when processing suboptimal images. To bridge this gap, we propose a robust report generation framework explicitly designed for image quality variations. We first introduce an Automated Quality Assessment Agent (AQAA) to identify low-quality samples within the MIMIC-CXR dataset and establish the Low-quality Radiology Report Generation (LRRG) benchmark. To tackle degradation-induced shifts, we propose a novel Dual-loop Training Strategy leveraging bi-level optimization and gradient consistency. This approach ensures the model learns quality-agnostic diagnostic features by aligning gradient directions across varying quality regimes. Extensive experiments demonstrate that our approach effectively mitigates model performance degradation caused by image quality deterioration. The code and data will be released upon acceptance.

cs.CV↗

arXiv2026

MedPruner: Training-Free Hierarchical Token Pruning for Efficient 3D Medical Image Understanding in Vision-Language Models

While specialized Medical Vision-Language Models (VLMs) have achieved remarkable success in interpreting 2D and 3D medical modalities, their deployment for 3D volumetric data remains constrained by significant computational inefficiencies. Current architectures typically suffer from massive anatomical redundancy due to the direct concatenation of consecutive 2D slices and lack the flexibility to handle heterogeneous information densities across different slices using fixed pruning ratios. To address these challenges, we propose MedPruner, a training-free and model-agnostic hierarchical token pruning framework specifically designed for efficient 3D medical image understanding. MedPruner introduces a two-stage mechanism: an Inter-slice Anchor-based Filtering module to eliminate slice-level temporal redundancy, followed by a Dynamic Information Nucleus Selection strategy that achieves adaptive token-level compression by quantifying cumulative attention weights. Extensive experiments on three 3D medical benchmarks and across three diverse medical VLMs reveal massive token redundancy in existing architectures. Notably, MedPruner enables models such as MedGemma-1.5 to maintain or even exceed their original performance while retaining fewer than 5\% of visual tokens, thereby reducing visual-token overhead and validating the necessity of dynamic token selection for practical clinical deployment. Our code is available at https://github.com/CUHK-AIM-Group/MedPruner.

cs.CV↗

arXiv2026

Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters

We introduce Step 3.5 Flash, a sparse Mixture-of-Experts (MoE) model that bridges frontier-level agentic intelligence and computational efficiency. We focus on what matters most when building agents: sharp reasoning and fast, reliable execution. Step 3.5 Flash pairs a 196B-parameter foundation with 11B active parameters for efficient inference. It is optimized with interleaved 3:1 sliding-window/full attention and Multi-Token Prediction (MTP-3) to reduce the latency and cost of multi-round agentic interactions. To reach frontier-level intelligence, we design a scalable reinforcement learning framework that combines verifiable signals with preference feedback, while remaining stable under large-scale off-policy training, enabling consistent self-improvement across mathematics, code, and tool use. Step 3.5 Flash demonstrates strong performance across agent, coding, and math tasks, achieving 85.4% on IMO-AnswerBench, 86.4% on LiveCodeBench-v6 (2024.08-2025.05), 88.2% on tau2-Bench, 69.0% on BrowseComp (with context management), and 51.0% on Terminal-Bench 2.0, comparable to frontier models such as GPT-5.2 xHigh and Gemini 3.0 Pro. By redefining the efficiency frontier, Step 3.5 Flash provides a high-density foundation for deploying sophisticated agents in real-world industrial environments.

cs.CL↗

arXiv2026

SpotAgent: Grounding Visual Geo-localization in Large Vision-Language Models through Agentic Reasoning

Large Vision-Language Models (LVLMs) have demonstrated strong reasoning capabilities in geo-localization, yet they often struggle in real-world scenarios where visual cues are sparse, long-tailed, and highly ambiguous. Previous approaches, bound by internal knowledge, often fail to provide verifiable results, yielding confident but ungrounded predictions when faced with confounded evidence. To address these challenges, we propose SpotAgent, a framework that formalizes geo-localization into an agentic reasoning process that leverages expert-level reasoning to synergize visual interpretation with tool-assisted verification. SpotAgent actively explores and verifies visual cues by leveraging external tools (e.g., web search, maps) through a ReAct diagram. We introduce a 3-stage post-training pipeline starting with a Supervised Fine-Tuning (SFT) stage for basic alignment, followed by an Agentic Cold Start phase utilizing high-quality trajectories synthesized via a Multi-Agent framework, aiming to instill tool-calling expertise. Subsequently, the model's reasoning capabilities are refined through Reinforcement Learning. We propose a Spatially-Aware Dynamic Filtering strategy to enhance the efficiency of the RL stage by prioritizing learnable samples based on spatial difficulty. Extensive experiments on standard benchmarks demonstrate that SpotAgent achieves state-of-the-art performance, effectively mitigating hallucinations while delivering precise and verifiable geo-localization.

cs.AI↗

arXiv2026

IFNSO: Iteration-Free Newton-Schulz Orthogonalization

The Newton-Schulz (NS) iteration has become a key technique for orthogonalization in optimizers such as Muon and for optimization on the Stiefel manifold. Despite its effectiveness, the conventional NS iteration incurs significant computational overhead due to repeated high-dimensional matrix multiplications. To overcome these limitations, we propose Iteration-Free Newton-Schulz Orthogonalization (IFNSO), a novel framework that consolidates the traditional iterative structure into a unified and Iteration-Free formulation. By analyzing the contribution of individual matrix powers, we streamline the process by removing insignificant terms and introducing a polynomial with learnable coefficients. These coefficients are optimized to ensure both superior computational efficiency and stable convergence. Extensive experiments demonstrate that IFNSO achieves superior performance compared to existing methods. Our code is available at: https://github.com/greekinRoma/Unified_Newton_Schulz_Orthogonalization.

cs.LG↗

arXiv2025

Step-DeepResearch Technical Report

As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.

cs.CL↗

arXiv2025

Supermassive Black Holes with High Accretion Rates in Active Galactic Nuclei. XV. Reverberation Mapping of Mg II Emission Lines

As the 15th paper in a series reporting on a large reverberation mapping (RM) campaign of super-Eddington accreting massive black holes (SEAMBHs) in active galactic nuclei (AGNs), we present the results of measurements of the Mg II lines in 18 SEAMBHs monitored spectroscopically from 2017 to 2024. Among these, the time lags of Mg II have been successfully determined for 8 of the 18 objects, thereby expanding the current Mg II RM sample, particularly at higher accretion rates. By incorporating measurements of the line widths, we determine the masses of their central supermassive black holes. Based on these new measurements, we update the relation between the Mg II radius and the monochromatic luminosity at 3000 $\mathring{\mathrm{A}}$ ($R_{\rm MgII}-L_{3000}$ relation), yielding a slope of $0.24 \pm 0.03$, which is slightly shallower than, yet still consistent with, previously reported values. Similar to the H$\beta$ lines, the Mg II time lags in SEAMBHs are shorter than those of AGNs with normal accretion rates at comparable luminosities. The deviation of AGNs from the best-fit $R_{\rm MgII}-L_{3000}$ relation shows a strong correlation with the accretion rate, while no significant correlation is found between the deviation and the flux ratio of UV iron to Mg II.

astro-ph.GA↗

arXiv2025

STGBD-Net: Spatio-temporal Gradient Basis Decomposition Network for Infrared Small Target Detection

A key challenge in infrared small target detection (IRSTD) is that weak target signal responses are easily obscured by strong background clutter, frequently resulting in missed detections. While traditional gradient-based methods attempt to capture fine details, their robustness is limited by the static fusion of multi-directional gradient features. In this paper, we rethink feature fusion from the perspective of Basis Decomposition Theory and propose a novel framework that reformulates the process into an explicit and adaptive decomposition-and-reconstruction paradigm. Specifically, we introduce the Basis Decomposition Module (BDM) and its specialized variant, the Gradient Decomposition Module (GDM) for IRSTD. GDMs treat the normalized gradient features as basis vectors to reconstruct a new feature, thereby maintaining detailed structures and highlighting infrared small targets. By integrating GDMs into a lightweight three-stage U-Net, we develop two unified architectures: the Spatial Gradient Basis Decomposition Network for single-frame detection and the Spatio-temporal Gradient Basis Decomposition Network for multi-frame scenarios. Extensive experiments demonstrate that our networks achieve state-of-the-art (SOTA) performance across multiple benchmarks, offering a superior balance between detection accuracy and computational efficiency. Our codes will be made public at: https://github.com/greekinRoma/IRSTD_HC_Platform.

cs.CV↗

arXiv2025

Learning to Compress: Unlocking the Potential of Large Language Models for Text Representation

Text representation plays a critical role in tasks like clustering, retrieval, and other downstream applications. With the emergence of large language models (LLMs), there is increasing interest in harnessing their capabilities for this purpose. However, most of the LLMs are inherently causal and optimized for next-token prediction, making them suboptimal for producing holistic representations. To address this, recent studies introduced pretext tasks to adapt LLMs for text representation. Most of these tasks, however, rely on token-level prediction objectives, such as the masked next-token prediction (MNTP) used in LLM2Vec. In this work, we explore the untapped potential of context compression as a pretext task for unsupervised adaptation of LLMs. During compression pre-training, the model learns to generate compact memory tokens, which substitute the whole context for downstream sequence prediction. Experiments demonstrate that a well-designed compression objective can significantly enhance LLM-based text representations, outperforming models trained with token-level pretext tasks. Further improvements through contrastive learning produce a strong representation model (LLM2Comp) that outperforms contemporary LLM-based text encoders on a wide range of tasks while being more sample-efficient, requiring significantly less training data. Code is available at https://github.com/longtaizi13579/LLM2Comp.

cs.CL↗

arXiv2025

Detection of unexpected leading delays in broad H{\beta} line reverberations in the quasar PHL 1092

Delayed reverberations of broad emission lines in response to optical continuum variations have been widely observed in active galactic nuclei (AGNs). They serve as a powerful tool for probing inner structures of AGNs and estimating the masses of supermassive black holes (SMBHs). The delays exhibit a strong correlation with approximately the square root of the optical luminosity - a relationship known as the "standard structure" of AGN broad-line regions (BLRs). Here, we report the discovery of leading delays in H{\beta} line reverberations (LDRs) in the quasar PHL 1092 preceding variations of the 5100 {\AA} continuum by 17-57 days, based on our eight-year continuous campaign of reverberation mapping of super Eddington AGNs. The LDRs suggest that the 5100 {\AA} continuum regions are so extensive that they are larger than the BLRs. This phenomenon not only fundamentally disrupts the well-established BLR size-luminosity relation but also violates the principle of causality. This unprecedented LDRs challenge the conventional methods for estimating SMBH mass as well as the standard model of AGNs. A preferred scenario to explain the LDRs is that the SMBH-disk contains a population of accreting stellar-mass black holes (sMBHs) as extra heating sources of the disk. Consequently, continuum regions of the disk are efficiently stretched so that the 5100 {\AA} regions exceed the BLRs, yielding the observed LDRs. Generally, sMBH activities there could provide new physics of AGN phenomena, which can be tested by LIGO, LISA/Tianqin and ET detections of gravitational waves from sMBH mergers.

astro-ph.GA↗

arXiv2025

Revisiting Entropy Regularization: Adaptive Coefficient Unlocks Its Potential for LLM Reinforcement Learning

Reasoning ability has become a defining capability of Large Language Models (LLMs), with Reinforcement Learning with Verifiable Rewards (RLVR) emerging as a key paradigm to enhance it. However, RLVR training often suffers from policy entropy collapse, where the policy becomes overly deterministic, hindering exploration and limiting reasoning performance. While entropy regularization is a common remedy, its effectiveness is highly sensitive to the fixed coefficient, making it unstable across tasks and models. In this work, we revisit entropy regularization in RLVR and argue that its potential has been largely underestimated. Our analysis shows that (i) tasks of varying difficulty demand distinct exploration intensities, and (ii) balanced exploration may require the policy entropy to be maintained within a moderate range below its initial level. Therefore, we propose Adaptive Entropy Regularization (AER)--a framework that dynamically balances exploration and exploitation via three components: difficulty-aware coefficient allocation, initial-anchored target entropy, and dynamic global coefficient adjustment. Experiments on multiple mathematical reasoning benchmarks show that AER consistently outperforms baselines, improving both reasoning accuracy and exploration capability.

cs.LG↗

arXiv2025

Random Policy Valuation is Enough for LLM Reasoning with Verifiable Rewards

RL with Verifiable Rewards (RLVR) has emerged as a promising paradigm for improving the reasoning abilities of large language models (LLMs). Current methods rely primarily on policy optimization frameworks like PPO and GRPO, which follow generalized policy iteration that alternates between evaluating the current policy's value and improving the policy based on evaluation. While effective, they often suffer from training instability and diversity collapse, requiring complex heuristic tricks and careful tuning. We observe that standard RLVR in math reasoning can be formalized as a specialized finite-horizon Markov Decision Process with deterministic state transitions, tree-structured dynamics, and binary terminal rewards. Though large in scale, the underlying structure is simpler than general-purpose control settings for which popular RL algorithms (e.g., PPO) were developed, suggesting that several sophisticated techniques in existing methods may be reduced or even omitted. Based on this insight, we prove a surprising result: the optimal action can be recovered from the Q-function of a fixed uniformly random policy, thereby bypassing the generalized policy iteration loop and its associated heuristics. We introduce Random Policy Valuation for Diverse Reasoning (ROVER) to translate this principle into a practical and scalable algorithm for LLM math reasoning, a minimalist yet highly effective RL method that samples actions from a softmax over these uniform-policy Q-values. ROVER preserves diversity throughout training, allowing sustained exploration of multiple valid pathways. Across multiple base models and standard math reasoning benchmarks, ROVER demonstrates superior performance in both \textbf{quality} (\textbf{+8.2} on pass@1, \textbf{+16.8} on pass@256) and \textbf{diversity} (\textbf{+17.6\%}), despite its radical simplification compared to strong, complicated existing methods.

cs.LG↗