arXiv subjects

Jiayi Chen

Publications and source records attributed to Jiayi Chen.

At least 73 records · Page 4Linked to original sources

arXiv2026

CL-bench: A Benchmark for Context Learning

Current language models (LMs) excel at reasoning over prompts using pre-trained knowledge. However, real-world tasks are far more complex and context-dependent: models must learn from task-specific context and leverage new knowledge beyond what is learned during pre-training to reason and resolve tasks. We term this capability context learning, a crucial ability that humans naturally possess but has been largely overlooked. To this end, we introduce CL-bench, a real-world benchmark consisting of 500 complex contexts, 1,899 tasks, and 31,607 verification rubrics, all crafted by experienced domain experts. Each task is designed such that the new content required to resolve it is contained within the corresponding context. Resolving tasks in CL-bench requires models to learn from the context, ranging from new domain-specific knowledge, rule systems, and complex procedures to laws derived from empirical data, all of which are absent from pre-training. This goes far beyond long-context tasks that primarily test retrieval or reading comprehension, and in-context learning tasks, where models learn simple task patterns via instructions and demonstrations. Our evaluations of ten frontier LMs find that models solve only 17.2% of tasks on average. Even the best-performing model, GPT-5.1, solves only 23.7%, revealing that LMs have yet to achieve effective context learning, which poses a critical bottleneck for tackling real-world, complex context-dependent tasks. CL-bench represents a step towards building LMs with this fundamental capability, making them more intelligent and advancing their deployment in real-world scenarios.

cs.CL↗

arXiv2026

ALD-Derived WO3-x Leads to Nearly Wake-Up-Free Ferroelectric Hf0.5Zr0.5O2 at Elevated Temperatures

Breaking the memory wall in advanced computing architectures will require complex 3D integration of emerging memory materials such as ferroelectrics-either within the back-end-of-line (BEOL) of CMOS front-end processes or through advanced 3D packaging technologies. Achieving this integration demands that memory materials exhibit high thermal resilience, with the capability to operate reliably at elevated temperatures such as 125C, due to the substantial heat generated by front-end transistors. However, silicon-compatible HfO2-based ferroelectrics tend to exhibit antiferroelectric-like behavior in this temperature range, accompanied by a more pronounced wake-up effect, posing significant challenges to their thermal reliability. Here, we report that by introducing a thin tungsten oxide (WO3-x) layer-known as an oxygen reservoir-and carefully tuning its oxygen content, ultra-thin Hf0.5Zr0.5O2 (5 nm) films can be made robust against the ferroelectric-to-antiferroelectric transition at elevated temperatures. This approach not only minimizes polarization loss in the pristine state but also effectively suppresses the wake-up effect, reducing the required wake-up cycles from 105 to only 10 at 125C- a qualifying temperature for back-end memory integrated with front-end logic, as defined by the JEDEC standard. First-principles density functional theory calculations reveal that WO3 enhances the stability of the ferroelectric orthorhombic phase at elevated temperatures by increasing the tetragonal-to-orthorhombic phase energy gap, and promoting favorable phonon mode evolution, thereby supporting o-phase formation under both thermodynamic and kinetic constraints.

cond-mat.mtrl-sci↗

arXiv2026

Analysis of Sensing in OFDM-based ISAC under the Influence of Sampling Jitter

To enable integrated sensing and communication (ISAC) in cellular networks, a wide range of additional requirements and challenges are either imposed or become more critical. One such impairment is sampling jitter (SJ), which arises due to imperfections in the sampling instants of the clocks of digital-to-analog converters (DACs) and analog-to-digital converters (ADCs). While SJ is already well studied for communication systems based on orthogonal frequency-division multiplexing (OFDM), which is expected to be the waveform of choice for most sixth-generation (6G) scenarios where ISAC could be possible, the implications of SJ on the OFDM-based radar sensing must still be thoroughly analyzed. Considering that phase-locked loop (PLL)-based oscillators are used to derive sampling clocks, which leads to colored SJ, i.e., SJ with non-flat power spectral density, this article analyzes the resulting distortion of the adopted digital constellation modulation and sensing performance in OFDM-based ISAC for both baseband (BB) and bandpass (BP) sampling strategies and different oversampling factors. For BB sampling, it is seen that SJ induces intercarrier interference (ICI), while for BP sampling, it causes carrier phase error and more severe ICI due to a phase noise-like effect at the digital intermediate frequency. Obtained results for a single-input single-output OFDM-based ISAC system with various OFDM signal parameterizations demonstrate that SJ-induced degradation becomes non-negligible for both BB and BP sampling only for root mean square (RMS) SJ values above 10^-11 s at both DAC and ADC, which corresponds to 0.5*10^-2 times the considered critical sampling period without oversampling. Based on the achieved results, it can be concluded that state-of-the-art hardware enables sufficient communication and sensing robustness against SJ, as RMS SJ values in the femtosecond range can be achieved.

eess.SP↗

arXiv2026

Double Hall-Littlewood symmetric polynomials

We establish a ring isomorphism between the derived Hall algebra of the Jordan quiver and the ring of double symmetric functions (i.e., the ring of symmetric polynomials in two sets of countably many variables, invariant under the respective actions of their symmetric groups) with a parameter $t$. This isomorphism maps the derived Hall basis (the natural basis of the derived Hall algebra) to a class of double Hall-Littlewood (HL) symmetric functions, which are formulated via raising and lowering operators. These double HL functions are parameterized by bipartitions; they reduce to the classical HL functions when one of the partitions is empty, and specialize to Schur Laurent symmetric functions at $t = 0$. We also derive the Pieri rules for these double HL functions. Additionally, we obtain several natural generating functions for the derived Hall algebra as well as their transition relations, which can be transferred to the ring of double symmetric functions via the established ring isomorphism.

math.QA↗

arXiv2026

iQuantum groups and iHopf algebras II: dual canonical bases

Building on the iHopf algebra realization of quasi-split universal iquantum groups developed in a prequel, we construct the dual canonical basis for a universal iquantum group of arbitrary finite type, which are further shown to be preserved by the ibraid group action; this recovers the results of Lu-Pan in ADE type obtained earlier in a geometric approach. Moreover, we identify the dual canonical basis for the Drinfeld double quantum group of arbitrary finite type, which is realized via iHopf algebra on the double Borel, with Berenstein-Greenstein's double canonical basis, settling several of their conjectures.

math.QA↗

arXiv2025

Hf/Zr Superlattice-Based High-\k{appa} Gate Dielectrics with Dipole Layer Engineering for Advanced CMOS

Advanced logic transistors require gate dielectrics that achieve sub-nanometer equivalent oxide thickness (EOT), suppress leakage, and satisfy three key requirements: (i) compatibility with RMG-like high-temperature processing, (ii) sufficient Vth tunability for multi-Vth design, and (iii) high device reliability. However, meeting all of these requirements simultaneously has been difficult with conventional high-k systems. In this work, we demonstrate that Hf/Zr-based gate stacks quantitatively satisfy these conditions. After a 700 C N2 anneal, the HZH superlattice achieves an EOT of 7.3 A, lower than conventional HfO2-only stacks (8.5 A) while maintaining comparable leakage. Embedding a 3 A Al2O3 dipole within an HfO2/ZrO2/HfO2 superlattice (HZHA) breaks the conventional dipole trade-off, achieving an EOT of 8.4 A, lower than the 9.0 A of a standard HfO2/Al2O3 stack, while providing a flatband voltage shift greater than 200 mV, thereby enabling multi-Vth tuning without compromising scaling. Furthermore, under -2 V negative-bias temperature stress at 125 C for 100 s, HZHA and HA exhibit comparable flatband voltage drifts of 87 mV and 97 mV, respectively, confirming that strong Vth tunability and sub-nanometer EOT can be achieved without compromising stability. In addition to these quantitative advances, this study reveals previously unreported physical insights into dipole behavior and interfacial diffusion in ultrathin Hf/Zr multilayers. These results establish HZHA as an RMG-compatible, Vth-tunable, low-EOT dielectric platform capable of supporting logic scaling beyond the 1 nm frontier.

cond-mat.mes-hall↗

arXiv2025

MARS: A Meta-Adaptive Reinforcement Learning Framework for Risk-Aware Multi-Agent Portfolio Management

Reinforcement Learning (RL) has shown significant promise in automated portfolio management; however, effectively balancing risk and return remains a central challenge, as many models fail to adapt to dynamically changing market conditions. We propose Meta-controlled Agents for a Risk-aware System (MARS), a novel framework addressing this through a multi-agent, risk-aware approach. MARS replaces monolithic models with a Heterogeneous Agent Ensemble, where each agent's unique risk profile is enforced by a Safety-Critic network to span behaviors from capital preservation to aggressive growth. A high-level Meta-Adaptive Controller (MAC) dynamically orchestrates this ensemble, shifting reliance between conservative and aggressive agents to minimize drawdown during downturns while seizing opportunities in bull markets. This two-tiered structure leverages behavioral diversity rather than explicit feature engineering to ensure a disciplined portfolio robust across market regimes. Experiments on major international indexes confirm that our framework significantly reduces maximum drawdown and volatility while maintaining competitive returns.

cs.LG↗

arXiv2025

R2Q: Towards Robust 2-Bit Large Language Models via Residual Refinement Quantization

The rapid progress of Large Language Models (LLMs) has brought substantial computational and memory demands, spurring the adoption of low-bit quantization. While 8-bit and 4-bit formats have become prevalent, extending quantization to 2 bits remains challenging due to severe accuracy degradation. To address this, we propose Residual Refinement Quantization (R2Q)-a novel 2-bit quantization framework that decomposes the process into two sequential 1-bit sub-quantizations, forming an adaptive quantization lattice. Extensive evaluations on Llama, OPT, and Qwen across diverse benchmarks-covering question answering, commonsense reasoning, and language modeling-demonstrate that R2Q consistently outperforms existing 2-bit quantization methods in both fine-grained and coarse-grained settings. By refining quantization through a residual learning mechanism, R2Q enhances performance, improves training stability, and accelerates convergence under extreme compression. Furthermore, its modular design enables seamless integration with existing quantization-aware training (QAT) frameworks.

cs.CL↗

arXiv2025

Higgs production via vector-boson fusion at the LHC

In this article, we summarise the recent experimental measurements and theoretical work on Higgs boson production via vector-boson fusion at the LHC. Along with this, we provide state-of-the-art predictions at fixed order as well as with parton-shower corrections within the Standard Model at 13.6 TeV. The results are presented in the form of multi-differential distributions as well as in the Simplified Template Cross Section bins. All materials and outputs of this study are available on public repositories. Finally, following findings in the literature, recommendations are made to estimate theoretical uncertainties related to parton-shower corrections.

hep-ph↗

arXiv2025

Curvature of high-dimensional data

We consider the problem of estimating curvature where the data can be viewed as a noisy sample from an underlying manifold. For manifolds of dimension greater than one there are multiple definitions of local curvature, each suggesting a different estimation process for a given data set. Recently, there has been progress in proving that estimates of ``local point cloud curvature" converge to the related smooth notion of local curvature as the density of the point cloud approaches infinity. Herein we investigate practical limitations of such convergence theorems and discuss the significant impact of bias in such estimates as reported in recent literature. We provide theoretical arguments for the fact that bias increases drastically in higher dimensions, so much so that in high dimensions, the probability that a naive curvature estimate lies in a small interval near the true curvature could be near zero. We present a probabilistic framework that enables the construction of more accurate estimators of curvature for arbitrary noise models. The efficacy of our technique is supported with experiments on spheres of dimension as large as twelve.

math.ST↗

arXiv2025

A Stylometric Application of Large Language Models

We show that large language models (LLMs) can be used to distinguish the writings of different authors. Specifically, an individual GPT-2 model, trained from scratch on the works of one author, will predict held-out text from that author more accurately than held-out text from other authors. We suggest that, in this way, a model trained on one author's works embodies the unique writing style of that author. We first demonstrate our approach on books written by eight different (known) authors. We also use this approach to confirm R. P. Thompson's authorship of the well-studied 15th book of the Oz series, originally attributed to F. L. Baum.

cs.CL↗

arXiv2025

EvaLearn: Quantifying the Learning Capability and Efficiency of LLMs via Sequential Problem Solving

We introduce EvaLearn, a pioneering benchmark designed to evaluate large language models (LLMs) on their learning capability and efficiency in challenging tasks, a critical, yet underexplored aspect of model potential. EvaLearn contains 648 challenging problems across six task types, grouped into 182 sequences, each sequence dedicated to one task type. Diverging from most existing benchmarks that evaluate models in parallel, EvaLearn requires models to solve problems sequentially, allowing them to leverage the experience gained from previous solutions. EvaLearn provides five comprehensive automated metrics to evaluate models and quantify their learning capability and efficiency. We extensively benchmark nine frontier models and observe varied performance profiles: some models, such as Claude-3.7-sonnet, start with moderate initial performance but exhibit strong learning ability, while some models struggle to benefit from experience and may even show negative transfer. Moreover, we investigate model performance under two learning settings and find that instance-level rubrics and teacher-model feedback further facilitate model learning. Importantly, we observe that current LLMs with stronger static abilities do not show a clear advantage in learning capability across all tasks, highlighting that EvaLearn evaluates a new dimension of model performance. We hope EvaLearn provides a novel evaluation perspective for assessing LLM potential and understanding the gap between models and human capabilities, promoting the development of deeper and more dynamic evaluation approaches. All datasets, the automatic evaluation framework, and the results studied in this paper are available at the GitHub repository.

cs.CL↗

arXiv2025

Evaluation of Quantum Offset in Velocity Imaging-Based Electron Spectrometry

Velocity-map imaging of electrons is a pivotal technique in chemical physics. A recent study reported a quantum offset as large as 0.2 cm-1 in velocity imaging-based electron spectrometry [Phys. Rev. Lett. 134, 043001 (2025)]. In this work, we assess the existence this offset through a combination of simulations and experiments. Our simulations reveal that the velocity imaging results reconstructed using the maximum entropy velocity Legendre reconstruction (MEVELER) method exhibit no such offset. Furthermore, experimental measurements of the electron affinity of oxygen conducted at various imaging voltages show no discernible offset attributable to the electric field in the photodetachment region. Therefore, we conclude that there is no evidence for the claimed quantum offset in properly analyzed velocity imaging-based electron spectrometry.

physics.atom-ph↗

arXiv2025

A Joint Learning Approach to Hardware Caching and Prefetching

Several learned policies have been proposed to replace heuristics for scheduling, caching, and other system components in modern systems. By leveraging diverse features, learning from historical trends, and predicting future behaviors, such models promise to keep pace with ever-increasing workload dynamism and continuous hardware evolution. However, policies trained in isolation may still achieve suboptimal performance when placed together. In this paper, we inspect one such instance in the domain of hardware caching -- for the policies of cache replacement and prefetching. We argue that these two policies are bidirectionally interdependent and make the case for training the two jointly. We propose a joint learning approach based on developing shared representations for the features used by the two policies. We present two approaches to develop these shared representations, one based on a joint encoder and another based on contrastive learning of the embeddings, and demonstrate promising preliminary results for both of these. Finally, we lay down an agenda for future research in this direction.

cs.LG↗

arXiv2025

Warm absorber outflows in radio-loud active galactic nucleus 3C~59

Both jets and ionized outflows in active galactic nuclei (AGNs) are thought to play important roles in affecting the star formation and evolution of host galaxies, but their relationship is still unclear. As a pilot study, we performed a detailed spectral analysis for a radio-loud (RL) AGN 3C~59 ($z=0.1096$) by systematically considering various factors that may affect the fitting results, and thereby establishing a general spectral fitting strategy for subsequent research with larger sample. 3C~59 is one rare target for simultaneously studying jets and warm absorbers (WAs) that is one type of ionized outflows. Based on the multi-wavelength data from near-infrared (NIR) to hard X-ray bands detected by DESI, GALEX, and XMM-Newton, we used SPEX code to build broadband continuum models and perform photoionization modeling with PION code to constrain the physical parameters of WAs in 3C~59. We found two WAs with ionization parameter of $\log [ξ/(\rm{erg\ cm\ s}^{-1})] = 2.65^{+0.10}_{-0.09}$ and $1.65\pm 0.11$, respectively, and their outflowing velocities are $v_{\rm out} = -528^{+163}_{-222}\ \rm{km\ s}^{-1}$ and $-228^{+121}_{-122}\ \rm{km\ s}^{-1}$, respectively. These WAs are located between outer torus and narrow (emission-)line region, and their positive $v_{\rm out}$-$ξ$ relation can be explained by the radiation-pressure-driven mechanism. We found that the estimations of these physical properties are affected by the different spectral fitting strategies, such as the inclusion of NIR to ultra-violet data, the choice of energy range of spectrum, or the composition of the spectral energy distribution. Based on the same fitting strategy, this work presents a comparative study of outflow driven mechanism between a RL AGN (3C 59) and a radio-quiet AGN (NGC 3227), which suggests a similar driven mechanism of their WA outflows and a negligible role of jets in this process.

astro-ph.GA↗

arXiv2025

FlowVLA: Visual Chain of Thought-based Motion Reasoning for Vision-Language-Action Models

Many Vision-Language-Action (VLA) models are built upon an internal world model trained via next-frame prediction ``$v_t \rightarrow v_{t+1}$''. However, this paradigm attempts to predict the future frame's appearance directly, without explicitly reasoning about the underlying dynamics. \textbf{This lack of an explicit motion reasoning step} often leads to physically implausible visual forecasts and inefficient policy learning. To address this limitation, we introduce the \textbf{Visual Chain of Thought (Visual CoT)}, a paradigm that compels the model to first reason about \textbf{motion dynamics} before generating the future frame. We instantiate this paradigm by proposing \textbf{FlowVLA}, an autoregressive Transformer that explicitly materializes this reasoning process as ``$v_t \rightarrow f_t \rightarrow v_{t+1}$'', where $f_t$ is an intermediate optical flow prediction that inherently encodes motion. By forcing the model to first follow the motion plan encoded by $f_t$, this process inherently \textbf{aligns the pre-training objective of dynamics prediction with the downstream task of action generation.} We conduct experiments on challenging robotics manipulation benchmarks, as well as real-robot evaluations. Our FlowVLA not only generates \textbf{more coherent and physically plausible visual predictions}, but also achieves state-of-the-art policy performance with \textbf{substantially improved sample efficiency}, pointing toward a more principled foundation for world modeling in VLAs. Project page: https://irpn-lab.github.io/FlowVLA/

cs.RO↗

arXiv2025

A purely Quantum Generative Modeling through Unitary Scrambling and Collapse

Quantum computing offers fundamentally more expressive mechanisms for generative modeling, yet current approaches remain constrained by classical neural components that bottleneck quantum capability and hardware efficiency. We propose the Quantum Scrambling and Collapse Generative Model (QGen), a purely quantum paradigm that eliminates classical dependencies. QGen implements two coherent processes: scrambling, which interleaves Gaussian diffusion channels with unitary delocalization to disperse information globally while avoiding collapse into uninformative states; and collapse, where parameterized quantum circuits refocus scrambled distributions into structured outputs, achieving distributional reconstruction under coherent evolution. To enable scalability, we introduce a measurement-based training principle that decomposes learning into tractable subproblems, mitigating barren plateaus. Empirically, QGen outperforms classical and hybrid baselines under matched parameter budget, while maintaining robustness under finite-shot sampling, demonstrating strong feasibility for near-term hardware.

quant-ph↗

arXiv2025

BODex: Scalable and Efficient Robotic Dexterous Grasp Synthesis Using Bilevel Optimization

Robotic dexterous grasping is important for interacting with the environment. To unleash the potential of data-driven models for dexterous grasping, a large-scale, high-quality dataset is essential. While gradient-based optimization offers a promising way for constructing such datasets, previous works suffer from limitations, such as inefficiency, strong assumptions in the grasp quality energy, or limited object sets for experiments. Moreover, the lack of a standard benchmark for comparing different methods and datasets hinders progress in this field. To address these challenges, we develop a highly efficient synthesis system and a comprehensive benchmark with MuJoCo for dexterous grasping. We formulate grasp synthesis as a bilevel optimization problem, combining a novel lower-level quadratic programming (QP) with an upper-level gradient descent process. By leveraging recent advances in CUDA-accelerated robotic libraries and GPU-based QP solvers, our system can parallelize thousands of grasps and synthesize over 49 grasps per second on a single 3090 GPU. Our synthesized grasps for Shadow, Allegro, and Leap hands all achieve a success rate above 75% in simulation, with a penetration depth under 1 mm, outperforming existing baselines on nearly all metrics. Compared to the previous large-scale dataset, DexGraspNet, our dataset significantly improves the performance of learning models, with a success rate from around 40% to 80% in simulation. Real-world testing of the trained model on the Shadow Hand achieves an 81% success rate across 20 diverse objects. The codes and datasets are released on our project page: https://pku-epic.github.io/BODex.

cs.RO↗