arXiv subjects

Jin Chen

Publications and source records attributed to Jin Chen.

At least 37 records · Page 2Linked to original sources

arXiv2026

A Causality-aware Infer-diagnose-refine Framework for Test-time Modality Adaptation in VLA Models

Vision-language-action (VLA) models predict sequential actions to execute tasks specified by language instructions, conditioned on visual observations and proprioceptive states. However, how to fuse modalities in VLA models remains an open problem, since robot manipulation involves dynamic phases, such as long-distance movements and close-range interactions, in which the importance of visual observations may vary over time. In this paper, we propose an infer-diagnose-refine (IDR) framework, a model-agnostic framework that can be integrated with diverse VLA architectures for refining action predictions at test time. IDR first infers actions under factual and counterfactual scenarios of visual observations, and then diagnoses the causal effects of visual observations as the estimated dynamic importance, which is finally used to refine the action predictions in a training-free manner. We further design a causality-aware action refiner to realize the IDR framework, including zero-padding interventions for inferring counterfactual actions, norm-based quantification for diagnosing causal effects, and gated residual fusion for refining actions. Extensive experiments on both simulation benchmarks and real-world tasks show improvements in overall performance across multiple VLA backbones, demonstrating the efficacy of dynamically adjusting visual importance at test time.

cs.RO↗

arXiv2026

An integrated super resolution THz 3D imaging system based on a linear nonlocal achromatic freeform Bessel beam lens and high power oscillator radiator array

High performance terahertz (THz) 3D imaging is critical for non-destructive evaluation. However, conventional architectures are fundamentally limited by severe chromatic aberrations, modest spatial resolution, restricted depths of focus (DOF), and the bulky nature of commercial transceivers. While metasurfaces offer a compact alternative, achieving broadband achromatic super-resolution with an extended DOF remains a formidable challenge. Here, we present a highly integrated 3D THz imaging platform that synergizes a 3D printed nonlocal freeform Bessel-beam lens with a high power, 65nm CMOS oscillator radiator array. Harnessing nonlocal interactions within the lens, we generate an achromatic super resolution Bessel beam (0.3 to 1 THz) with a subdiffraction full width at half maximum (FWHM) of 0.65λ and a robust 4.7-mm DOF. Crucially, the system overcomes conventional sidelobe limitations, enabling high-fidelity 2D imaging of intricate sub-millimeter targets (e.g., USAF 1951 charts and QR codes) alongside robust 3D volumetric imaging through highly scattering media, such as printed circuit boards. By converging standard CMOS technology with additive manufacturing, this work establishes a versatile, cost-effective paradigm for next-generation integrated THz photonics

physics.optics↗

arXiv2026

Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields

Recent years have witnessed the rapid evolution of AI agents toward handling increasingly complex, real-world tasks. However, existing benchmarks rarely evaluate whether agents can operate graphical user interfaces to complete long-horizon, high-value professional workflows across diverse domains. Current GUI benchmarks still predominantly focus on general-purpose software, relatively simple applications, and short-horizon tasks, leaving it largely unknown whether modern agents can follow user instructions to autonomously operate domain-specific professional software and accomplish economically valuable work in an end-to-end manner. To bridge this gap, we introduce Workflow-GYM, a benchmark for long-horizon GUI tasks centered on professional domains and specialized software environments. Through extensive experiments on state-of-the-art models, we find that even the strongest models achieve only slightly above 30% success rates, highlighting that professional long-horizon GUI workflows remain highly challenging for current GUI agents. Further analysis reveals that current agents struggle to maintain long-horizon workflow consistency, frequently exhibiting workflow stage omission, error propagation, objective drift, and insufficient understanding of professional software environments. Our findings provide important insights into the limitations of current agent systems and suggest key directions for the next generation of GUI-agent research.

cs.AI↗

arXiv2026

MSQA: A Natively Sourced Multilingual and Multicultural SimpleQA Benchmark

Multilingual fluency often invites a stronger assumption: a model that can speak a user's language must also understand the culture encoded by that language. We call this the Illusion of Cultural Alignment. To test this assumption directly, we introduce MSQA, a benchmark of 1,064 natively sourced questions across 11 language groups, five cultural dimensions, and three difficulty tiers. Unlike translated benchmarks, MSQA targets locally grounded knowledge and reduces shortcuts from English-centric cross-lingual transfer. Evaluating 18 LLMs, we find substantial cultural degradation and a pronounced Locality Effect: cultural competence tracks pre-training exposure more closely than general reasoning ability. We further show that common inference-time remedies do not dissolve the illusion. Models remain overconfident on unfamiliar cultural questions, repeated sampling yields unstable rather than reliable correctness, and retrieval augmentation helps unevenly on long-tail facts. These findings indicate that cultural alignment cannot be inferred from multilingual ability alone and requires deeper intervention than calibration, sampling, or retrieval at inference time

cs.CL↗

arXiv2026

Is Diversity All You Need for Scalable Robotic Manipulation?

Data scaling has driven remarkable success in foundation models for Natural Language Processing (NLP) and Computer Vision (CV), yet the principles of effective data scaling in robotic manipulation remain insufficiently understood. In this work, we investigate the nuanced role of data diversity in robot learning by examining three critical dimensions-task (what to do), embodiment (which robot to use), and expert (who demonstrates)-challenging the conventional intuition of "more diverse is better". Throughout extensive experiments on various robot platforms, we reveal that (1) task diversity proves more critical than per-task demonstration quantity, benefiting transfer from diverse pre-training tasks to novel downstream scenarios; (2) multi-embodiment pre-training data is optional for cross-embodiment transfer-models trained on high-quality single-embodiment data can efficiently transfer to different platforms, showing more desirable scaling property during fine-tuning than multi-embodiment pre-trained models; and (3) expert diversity, arising from individual operational preferences and stochastic variations in human demonstrations, can be confounding to policy learning, with velocity multimodality emerging as a key contributing factor. Based on this insight, we propose a distribution debiasing method to mitigate velocity ambiguity, the yielding GO-1-Pro achieves substantial performance gains of 15%, equivalent to using 2.5 times pre-training data. Collectively, these findings provide new perspectives and offer practical guidance on how to scale robotic manipulation datasets effectively.

cs.RO↗

arXiv2026

EgoHumanoid: Unlocking In-the-Wild Loco-Manipulation with Robot-Free Egocentric Demonstration

Human demonstrations offer rich environmental diversity and scale naturally, making them an appealing alternative to robot teleoperation. While this paradigm has advanced robot-arm manipulation, its potential for the more challenging, data-hungry problem of humanoid loco-manipulation remains largely unexplored. We present EgoHumanoid, the first framework to co-train a vision-language-action policy using abundant egocentric human demonstrations together with a limited amount of robot data, enabling humanoids to perform loco-manipulation across diverse real-world environments. To bridge the embodiment gap between humans and robots, including discrepancies in physical morphology and viewpoint, we introduce a systematic alignment pipeline spanning from hardware design to data processing. A portable system for scalable human data collection is developed, and we establish practical collection protocols to improve transferability. At the core of our human-to-humanoid alignment pipeline lies two key components. The view alignment reduces visual domain discrepancies caused by camera height and perspective variation. The action alignment maps human motions into a unified, kinematically feasible action space for humanoid control. Extensive real-world experiments demonstrate that incorporating robot-free egocentric data significantly outperforms robot-only baselines by 51\%, particularly in unseen environments. Our analysis further reveals which behaviors transfer effectively and the potential for scaling human data.

cs.RO↗

arXiv2026

Expand More, Shrink Less: Shaping Effective-Rank Dynamics for Dense Scaling in Recommendation

Scaling recommendation models is a central challenge in recommender systems. Recently, RankMixer has emerged as an effective solution, operating on a unified token representation and alternating between token mixing and per-token feedforward networks (P-FFNs) to achieve scalable performance. However, RankMixer suffers from \textit{embedding collapse}, where learned representations have low effective rank, limiting expressivity and underutilizing the expanded representation space. Through empirical analysis and theoretical insights, we identify rigid token mixing and P-FFN modules as the primary causes of this phenomenon, jointly inducing a \textbf{damped oscillatory trajectory} in effective-rank evolution across layers. To address it, we propose RankElastor, a novel architecture that produces spectrum-robust representations with provable collapse mitigation. RankElastor introduces two components: (i) \textbf{parameterized full mixing}, which enables expressive token mixing with improved spectral robustness; and (ii) \textbf{GLU-improved P-FFNs}, which stabilize representation spectra through GLU-style FFN modules. Extensive experiments on large-scale industrial datasets demonstrate that RankElastor consistently improves recommendation performance, mitigates embedding collapse, and exhibits robust scaling behavior. Code is available at this GitHub repository: https://github.com/vasile-paskardlgm/RankElastor

cs.LG↗

arXiv2026

Fast solvers for Tokamak fluid models with PETSC

Multigrid (MG) is widely recognized as a highly effective solver for the model problem, the Laplacian, but textbook MG fails on most problems of interest. MG methods have been applied to complex, real-world applications with careful consideration of the physical model and discretization. This work develops the first step in applying MG methods to science and engineering relevant magnetohydrodynamics (MHD) tokamak models in the \textit{M3D-C1} https://m3dc1.pppl.gov fusion energy science code. The semi-implicit time integrator in \textit{M3D-C1} is composed of many linear solves. The implicit advance of the momentum equation is the most challenging and is the focus of this work. The current production solver in \textit{M3D-C1} is a block Jacobi (BJ) preconditioner within a Krylov solver, where blocks group degrees of freedom on planes of constant toroidal coordinate. BJ convergence degrades as the number of planes increases due to the spectral properties of the matrix preconditioned with BJ. The partially magnetic field-aligned, regular toroidal grid structure in \textit{M3D-C1} is amenable to semi-coarsening geometric MG in the toroidal direction. This paper develops such a solver and demonstrates competitive performance on a runaway electron model of a SPARC https://cfs.energy/technology/sparc disruption, and superior robustness on a stellarator model on which the BJ solver fails to converge.

physics.plasm-ph↗

arXiv2026

RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems

The scaling laws for recommender systems have been increasingly validated, where MetaFormer-based architectures consistently benefit from increased model depth, hidden dimensionality, and user behavior sequence length. However, whether representation capacity scales proportionally with parameter growth remains unexplored. Prior studies on RankMixer reveal that the effective rank of token representations exhibits a damped oscillatory trajectory across layers, failing to increase consistently with depth and even degrading in deeper layers. Motivated by this observation, we propose RankUp, an architecture designed to mitigate representation collapse and enhance expressive capacity through randomized permutation splitting over sparse features, a multi-embedding paradigm, global token integration and crossed pretrained embedding tokens. RankUp has been fully deployed in large-scale production across Weixin Video Accounts, Official Accounts and Moments, yielding GMV improvements of 3.41%, 4.81% and 2.12%, respectively.

cs.IR↗

arXiv2026

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4.

cs.CL↗

arXiv2026

Classification of 2D Fermionic Systems with a $\mathbb Z_2$ Flavor Symmetry

We classify superfusion categories describing two-dimensional fermionic systems equipped with the universal fermion-parity symmetry, implemented by a topological defect line (TDL) $Z$, and an additional $\mathbb{Z}_2$ flavor symmetry generated by a $W$ TDL. Depending on whether $W$ is m-type or q-type, its fusion rules lead to three distinct classes, and solving the super-pentagon equations yields 16 consistent superfusion categories. These are labeled by invariants $(ν_W,ν_Z,ν_{WZ})$, which determine the $\mathbb{Z}_8$ anomaly classes of the symmetries generated by $W$, $Z$, and $WZ$. We also provide explicit realizations using multiple Majorana fermions and comment on implications for fermionic CFTs and gapped phases.

hep-th↗

arXiv2026

LogicPoison: Logical Attacks on Graph Retrieval-Augmented Generation

Graph-based Retrieval-Augmented Generation (GraphRAG) enhances the reasoning capabilities of Large Language Models (LLMs) by grounding their responses in structured knowledge graphs. Leveraging community detection and relation filtering techniques, GraphRAG systems demonstrate inherent resistance to traditional RAG attacks, such as text poisoning and prompt injection. However, in this paper, we find that the security of GraphRAG systems fundamentally relies on the topological integrity of the underlying graph, which can be undermined by implicitly corrupting the logical connections, without altering surface-level text semantics. To exploit this vulnerability, we propose \textsc{LogicPoison}, a novel attack framework that targets logical reasoning rather than injecting false contents. Specifically, \textsc{LogicPoison} employs a type-preserving entity swapping mechanism to perturb both global logic hubs for disrupting overall graph connectivity and query-specific reasoning bridges for severing essential multi-hop inference paths. This approach effectively reroutes valid reasoning into dead ends while maintaining surface-level textual plausibility. Comprehensive experiments across multiple benchmarks demonstrate that \textsc{LogicPoison} successfully bypasses GraphRAG's defenses, significantly degrading performance and outperforming state-of-the-art baselines in both effectiveness and stealth. Our code is available at \textcolor{blue}https://github.com/Jord8061/logicPoison.

cs.CL↗

arXiv2026

The Language of Touch: Translating Vibrations into Text with Dual-Branch Learning

The standardization of vibrotactile data by IEEE P1918.1 workgroup has greatly advanced its applications in virtual reality, human-computer interaction and embodied artificial intelligence. Despite these efforts, the semantic interpretation and understanding of vibrotactile signals remain an unresolved challenge. In this paper, we make the first attempt to address vibrotactile captioning, {\it i.e.}, generating natural language descriptions from vibrotactile signals. We propose Vibrotactile Periodic-Aperiodic Captioning (ViPAC), a method designed to handle the intrinsic properties of vibrotactile data, including hybrid periodic-aperiodic structures and the lack of spatial semantics. Specifically, ViPAC employs a dual-branch strategy to disentangle periodic and aperiodic components, combined with a dynamic fusion mechanism that adaptively integrates signal features. It also introduces an orthogonality constraint and weighting regularization to ensure feature complementarity and fusion consistency. Additionally, we construct LMT108-CAP, the first vibrotactile-text paired dataset, using GPT-4o to generate five constrained captions per surface image from the popular LMT-108 dataset. Experiments show that ViPAC significantly outperforms the baseline methods adapted from audio and image captioning, achieving superior lexical fidelity and semantic alignment.

cs.CV↗

arXiv2026

$χ_{0}$: Resource-Aware Robust Manipulation via Taming Distributional Inconsistencies

High-reliability long-horizon robotic manipulation has traditionally relied on large-scale data and compute to understand complex real-world dynamics. However, we identify that the primary bottleneck to real-world robustness is not resource scale alone, but the distributional shift among the human demonstration distribution, the inductive bias learned by the policy, and the test-time execution distribution -- a systematic inconsistency that causes compounding errors in multi-stage tasks. To mitigate these inconsistencies, we propose $χ_{0}$, a resource-efficient framework with effective modules designated to achieve production-level robustness in robotic manipulation. Our approach builds off three technical pillars: (i) Model Arithmetic, a weight-space merging strategy that efficiently soaks up diverse distributions of different demonstrations, varying from object appearance to state variations; (ii) Stage Advantage, a stage-aware advantage estimator that provides stable, dense progress signals, overcoming the numerical instability of prior non-stage approaches; and (iii) Train-Deploy Alignment, which bridges the distribution gap via spatio-temporal augmentation, heuristic DAgger corrections, and temporal chunk-wise smoothing. $χ_{0}$ enables two sets of dual-arm robots to collaboratively orchestrate long-horizon garment manipulation, spanning tasks from flattening, folding, to hanging different clothes. Our method exhibits high-reliability autonomy; we are able to run the system from arbitrary initial state for consecutive 24 hours non-stop. Experiments validate that $χ_{0}$ surpasses the state-of-the-art $π_{0.5}$ in success rate by nearly 250%, with only 20-hour data and 8 A100 GPUs. Code, data and models will be released to facilitate the community.

cs.RO↗

arXiv2026

Tunable Input-to-State Safety with Input Constraints

Tunable input-to-state safety (TISSf) generalizes the input-to-state safety (ISSf) framework by incorporating a tuning function that regulates safety conservatism while preserving robustness against perturbations. Despite its flexibility, the TISSf tuning function is often designed without explicitly incorporating actuator limits, which can lead to incompatibility with input constraints. To address this gap, this paper proposes a framework that integrates general compact input constraints into tuning function synthesis. Leveraging a geometric perspective, we characterize the TISSf condition as a state-dependent half-space constraint and derive a verifiable certificate for input compatibility using support functions. This characterization transforms the compatibility requirement into a design constraint on the tuning function, yielding a prescriptive lower bound that defines an admissible family of tunings under input constraints. These results are specialized to norm-bounded, polyhedral, and box constraints, yielding tractable control design conditions. We show that these conditions, combined with tuning function monotonicity, guarantee input compatibility and recursive feasibility of the resulting quadratic program (QP)-based safety filter. Furthermore, an offline parameter selection procedure using a covering-based sampling strategy ensures compatibility across the entire safe set via a linear program (LP). A connected cruise control (CCC) application demonstrates robust safety under TISSf while enforcing input constraints by design.

eess.SY↗

arXiv2026

Reexamining the strange metal charge response with transmission inelastic electron scattering

The strange metal remains one of the great unsolved problems for 21st century science. Since the early development of the marginal Fermi liquid phenomenology, it has been clear that progress requires detailed knowledge of the momentum- and frequency-dependent charge susceptibility, $χ(\mathbf{q},ω)$, particularly at large momenta. Electron energy-loss spectroscopy (EELS), performed in either reflection or transmission geometry, provides the most direct probe of $χ(\mathbf{q},ω)$. However, measurements over the past four decades have yielded conflicting results, with some studies reporting a dispersing RPA-like plasmon and others observing a strongly overdamped, incoherent response. Here we report a transmission EELS study of Bi$_2$Sr$_2$CaCu$_2$O$_{8+x}$ (Bi-2212) that simultaneously achieves high energy resolution ($ΔE \approx 30$ meV) and high momentum resolution ($Δq \approx 0.01$ Å$^{-1}$). To address issues of reproducibility, measurements were repeated ten times on five different Bi-2212 flakes, benchmarked against aluminum, a well-characterized Fermi liquid, and quantitatively compared with prior studies spanning four decades. At momenta $q < 0.15$ Å$^{-1}$, we observe a highly damped plasmon whose linewidth is comparable to its energy. At larger momenta, $q > 0.15$ Å$^{-1}$, this excitation does not disperse but instead evolves into an incoherent continuum, with no evidence for the RPA-like dispersion reported in some earlier works. Comparison with recent RIXS measurements on Bi-based cuprates supports the view that Bi-2212 is an incoherent metal with strongly damped charge excitations.

cond-mat.str-el↗

arXiv2026

BABE: Biology Arena BEnchmark

The rapid evolution of large language models (LLMs) has expanded their capabilities from basic dialogue to advanced scientific reasoning. However, existing benchmarks in biology often fail to assess a critical skill required of researchers: the ability to integrate experimental results with contextual knowledge to derive meaningful conclusions. To address this gap, we introduce BABE(Biology Arena BEnchmark), a comprehensive benchmark designed to evaluate the experimental reasoning capabilities of biological AI systems. BABE is uniquely constructed from peer-reviewed research papers and real-world biological studies, ensuring that tasks reflect the complexity and interdisciplinary nature of actual scientific inquiry. BABE challenges models to perform causal reasoning and cross-scale inference. Our benchmark provides a robust framework for assessing how well AI systems can reason like practicing scientists, offering a more authentic measure of their potential to contribute to biological research.

cs.AI↗

arXiv2026

Retrieval-Infused Reasoning Sandbox: A Benchmark for Decoupling Retrieval and Reasoning Capabilities

Despite strong performance on existing benchmarks, it remains unclear whether large language models can reason over genuinely novel scientific information. Most evaluations score end-to-end RAG pipelines, where reasoning is confounded with retrieval and toolchain choices, and the signal is further contaminated by parametric memorization and open-web volatility. We introduce DeR2, a controlled deep-research sandbox that isolates document-grounded reasoning while preserving core difficulties of deep search: multi-step synthesis, denoising, and evidence-based conclusion making. DeR2 decouples evidence access from reasoning via four regimes--Instruction-only, Concepts (gold concepts without documents), Related-only (only relevant documents), and Full-set (relevant documents plus topically related distractors)--yielding interpretable regime gaps that operationalize retrieval loss vs. reasoning loss and enable fine-grained error attribution. To prevent parametric leakage, we apply a two-phase validation that requires parametric failure without evidence while ensuring oracle-concept solvability. To ensure reproducibility, each instance provides a frozen document library (drawn from 2023-2025 theoretical papers) with expert-annotated concepts and validated rationales. Experiments across a diverse set of state-of-the-art foundation models reveal substantial variation and significant headroom: some models exhibit mode-switch fragility, performing worse with the Full-set than with Instruction-only, while others show structural concept misuse, correctly naming concepts but failing to execute them as procedures.

cs.AI↗