arXiv subjects

Yi Zheng

Publications and source records attributed to Yi Zheng.

At least 37 records · Page 2Linked to original sources

arXiv2026

Synthesis of single-layered fluorographdiyne nanosheets via selective on-surface 2D covalent polymerization

Two-dimensional conjugated polymers (2DCPs) are significant macromolecular materials with intriguing and tunable physicochemical properties that depend on their geometries. Graphdiyne and its derivatives are exemplary 2DCPs featuring sp-sp2 hybridized skeletons. However, achieving single-layered, large-domain/regular graphdiyne and its derivatives on surfaces remains a formidable challenge due to the lack of selective 2D covalent polymerization methods. Here, we report a selective on-surface 2D covalent polymerization method via the combination of cobalt catalysis and coronene templating, achieving the synthesis of single-layered fluorographdiyne nanosheets up to 60*60 nm2 on Au(111) surface. Using scanning probe techniques, we visualize the sequential polymerization process and characterize cobalt-activated coupling intermediates at the atomic level. Experimental and theoretical analyses suggest that strong d-π coupling between cobalt and alkynyl transforms a robust Csp-Au bond into a weaker Csp2-Au bond, thereby facilitating the demetallization C-C coupling. Besides, the templating effect of coronene suppresses kinetically trapped defects and improves the selectivity of hexagonal-ring formation in the complex 2D covalent polymerization process.

cond-mat.mtrl-sci↗

arXiv2026

Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models

Supervised fine-tuning (SFT) followed by reinforcement learning (RL) has become a standard post-training paradigm for large language models. This paradigm provides a cold-start for RL exploration, avoiding the inefficiency of pure RL where on-policy sampling yields insufficient positive samples. However, in practice, existing approaches often use a small amount of data for SFT initialization compared to the RL phase, which can cause the model to fit the limited samples and shift away from its pre-trained distribution. This distribution shift impedes the model's ability to effectively explore during subsequent RL training. To address this challenge, we propose that in low-data regimes, SFT should prioritize activating task-relevant capabilities rather than memorizing specific content. Along this line, we propose EKSFT (Entropy-KL Selective Fine-Tuning), which selectively masks tokens that exhibit either high entropy or high KL divergence from a reference model. By excluding these high-uncertainty, distribution-shifting tokens from imitation, EKSFT injects task-specific knowledge while preserving the integrity of the model's pre-trained distribution. Empirical evaluations on mathematical reasoning benchmarks demonstrate that EKSFT consistently outperforms standard SFT. Further RL fine-tuning from the EKSFT model yields consistently better post-RL performance, indicating improved exploration for the RL stage. Our codes and datasets are available at https://github.com/MINE-USTC/EKSFT.

cs.AI↗

arXiv2026

AgenticVBench: Can AI Agents Complete Real-World Post-Production Tasks?

Video production workflows offer a rich and demanding arena for evaluating multimodal AI agents: they require composite capabilities across text, image, audio, and video understanding, along with long-horizon planning, and tool use. To this end, we introduce AgenticVBench, a benchmark of 100 agentic tasks across 4 task families spanning the real world post-production workflow, constructed from real production workflows contributed by 20 industry experts averaging 6 years of professional experience. Tasks are paired with evaluation specifications that combine programmatic verifiers and expert rubrics. We evaluate frontier vision-language models (VLMs) with both vendor-native and open-source harnesses. The best evaluated agent stack barely crosses 30%, far below human expert performance on the same tasks. We further find that the choice of harness substantially affects model behavior, including scores, tool-use patterns, and failure modes. AgenticVBench provides a foundation for diagnosing and improving both models and harnesses for agentic video production. Benchmark website: https://agenticvbench.com.

cs.CR↗

arXiv2026

Cross-Subject EEG Emotion Recognition Based on Temporal Asynchronous Alignment Contrastive Learning

With the advancement of science and technology, the importance of emotion research has become increasingly evident. Electroencephalography (EEG)-based emotion recognition has emerged as an active research area in recent years, owing to its objectivity and high temporal resolution. However, most existing methods focus on optimizing encoder structures to enhance feature extraction capabilities, while paying relatively little attention to similarity calculation strategies, particularly overlooking the potential temporal misalignment of responses among different subjects. To address these shortcomings, this paper draws inspiration from the late interaction mechanism of ColBERT in natural language processing (NLP) and proposes a Temporal Asynchronous Alignment-based Contrastive Learning (TA2CL) framework. This method transforms the traditional global "hard alignment" similarity calculation approach into a fine-grained local matching mechanism, enabling the model to adaptively search for and align "locally highly correlated" segments between two EEG signals, thereby effectively mitigating the effects of inter-subject differences and temporal delays. Experimental results demonstrate that the proposed method achieves strong performance across multiple public datasets. Specifically, on the FACED dataset, it achieves an accuracy of 64.5% for the nine-class classification task and 79.5% for the binary classification task, while on the SEED and SEED-V datasets, it achieves accuracies of 86.4% and 70.1%, respectively, validating the method's effectiveness and generalization capability.

cs.HC↗

arXiv2026

Toward Natural and Companionable Virtual Agents via Cross-Temporal Emotional Modeling

Recent advances in foundation models have enabled conversational agents that aim for sustained companionship rather than mere task completion. Yet most still remain unable to support natural, long-term companion-like interactions, resulting in experiences that feel episodic and inauthentic. We argue that current agents overlooked cross-temporal modeling of agents' social behaviors and internal emotions: generated behaviors rarely influence an agent's emotional state, and emotional states seldom shape subsequent behaviors. We present Cross-Temporal Emotion Modeling (CTEM), a framework that links long-term behavioral history to moment-to-moment emotional expression. CTEM establishes a closed loop where past experiences update an evolving emotional state; this state conditions immediate interactions; and user feedback continually revises both memory and emotional state, enabling reflection and anticipation. We instantiate CTEM as Auri, a companion agent on an instant-messaging platform, and report a 21-day in-the-wild study showing that CTEM shows improvements in perceived naturalness, coherence, and emotional harmony.

cs.HC↗

arXiv2026

Scalable Generation of Massive Schrödinger Cat States via Quantum Tunneling

Massive objects in spatial superposition may provide insights into the interplay between quantum mechanics and gravity. Cold atomic interferometers offer a promising platform due to extended matter-wave coherence times and precise controllability. However, high-mass spatial superpositions beyond single atoms have yet to be generated in such setups. Here, we report the scalable realization of high-mass spatial entanglement via quantum tunneling of ultracold atoms in optical lattices. We observe coherent tunneling of bound clusters, forming a composite object with a mass of 608~amu. Full control of the model parameters allows us to mitigate the usual suppression of tunneling with increasing mass. Furthermore, we construct an interferometer to certify the entanglement and use spatially distributed Schrödinger cat states to perform quantum-enhanced measurements. These results establish an approach to generating and detecting massive superposition states relevant to studies of quantum gravity.

cond-mat.quant-gas↗

arXiv2026

PRA-RAG: Provably Robust Aggregation in Retrieval-Augmented Generation against Retrieval Corruption

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, effectively mitigating their inherent knowledge limitations. However, RAG remains vulnerable to poisoning attacks that manipulate retrieved texts to mislead model outputs. Existing defense mechanisms often lack theoretical robustness guarantees and perform unreliably when the LLM has limited knowledge of the retrieved content. In this work, we propose PRA-RAG, a provably robust retrieval aggregation algorithm designed to defend against poisoning attacks on retrieved texts. PRA-RAG samples multiple combinations of retrieved texts and utilizes geometric structures in the embedding space to identify a robust subset, from which a stable aggregated representation is derived. We provide theoretical bounds on the maximum impact of poisoned retrieved content and establish a quantitative measure of RAG's robustness. Experiments across multiple benchmarks and RAG architectures demonstrate that PRA-RAG reduces the attack success rate to as low as 1% while maintaining an accuracy of 71%, significantly outperforming representative state-of-the-art methods.

cs.IR↗

arXiv2026

Constraints on Halo Gas Profiles from Joint kSZ and Galaxy Clustering Analysis of ACT DR6 and CMASS

We measure the kinetic Sunyaev-Zel'dovich (kSZ) signal through a joint analysis of the pairwise kSZ effect and galaxy clustering using CMASS galaxies and ACT DR6 maps. This approach breaks degeneracies between the optical depth and nuisance parameters, enabling a reconstruction of the halo optical depth profile as a function of aperture scale. The kSZ signal reaches its highest signal-to-noise ratio of 7.2 at an aperture radius of $θ_{\rm AP} = 2$ arcmin, while the full profile rejects the no-kSZ hypothesis at $8.7σ$. Applying the same analysis pipeline to the Websky simulation, we find that the observed optical depth profile is somewhat more extended than the simulated one. This difference suggests that baryonic feedback in the real Universe may be stronger and redistribute gas to larger radii more efficiently than modeled in the simulation, although residual systematic effects and modeling uncertainties remain to be further investigated.

astro-ph.CO↗

arXiv2026

Thermally accessible broadband soliton microcombs in silicon carbide enabled by dynamic polarization control

Optical microcombs generated in high-Q microresonators are promising chip-scale light sources for applications ranging from optical communications to spectroscopy and metrology. However, thermo-optic instabilities remain a major obstacle to reliable soliton access. Self-cooling using auxiliary modes can stabilize the intracavity power, yet part of the power is continuously allocated to thermal compensation rather than comb generation, thereby limiting comb power and bandwidth. Here we propose a thermal compensation scheme based on dynamic polarization control. During soliton initiation, a fraction of the pump is coupled to an orthogonally polarized mode to provide self-cooling and ensure reliable soliton access. After soliton formation, polarization rotation and pump tuning transfer this cooling power to the comb-generating mode, enabling efficient single-soliton operation. Using this approach, we experimentally demonstrate a broadband 108-GHz-FSR single-soliton microcomb spanning over 450 nm, together with approximately 39% improvement in the 20-dB bandwidth and 60% increase in comb power relative to the static self-cooling configuration. This dynamic polarization-based thermal compensation enables efficient use of available laser power and provides a practical route to high-performance soliton microcombs in platforms with strong thermo-optic effects.

physics.optics↗

arXiv2026

Human-Centric Topic Modeling with Goal-Prompted Contrastive Learning and Optimal Transport

Existing topic modeling methods, from LDA to recent neural and LLM-based approaches, which focus mainly on statistical coherence, often produce redundant or off-target topics that miss the user's underlying intent. We introduce Human-centric Topic Modeling, \emph{Human-TM}), a novel task formulation that integrates a human-provided goal directly into the topic modeling process to produce interpretable, diverse and goal-oriented topics. To tackle this challenge, we propose the \textbf{G}oal-prompted \textbf{C}ontrastive \textbf{T}opic \textbf{M}odel with \textbf{O}ptimal \textbf{T}ransport (GCTM-OT), which first uses LLM-based prompting to extract goal candidates from documents, then incorporates these into semantic-aware contrastive learning via optimal transport for topic discovery. Experimental results on three public subreddit datasets show that GCTM-OT outperforms state-of-the-art baselines in topic coherence and diversity while significantly improving alignment with human-provided goals, paving the way for more human-centric topic discovery systems.

cs.AI↗

arXiv2026

Broadband parametric amplification in AlGaAs-on-insulator nanowaveguides

Optical amplification is critical for optical signal transmission. While the emergence of erbium-doped fiber amplifiers has revolutionized optical communications in fiber-based systems, on-chip amplification remains essential for integrated optics. Nanoscale waveguides enhance nonlinearity by several orders of magnitude, making them promising candidates for optical parametric amplification. Using a pulsed pump at 1550 nm, broadband optical parametric amplification based on four-wave mixing is investigated in AlGaAs-on-insulator nanowaveguides. The strong nonlinearity enables an on-off gain as high as 58.4 dB. Meanwhile, the low propagation loss leads to a net on-chip gain of 56.2 dB. With further dispersion engineering, the net on-chip gain bandwidth extends beyond 415 nm, which is 2.3 times larger than previous reports pumped in the telecom band in integrated optics. These results represent the largest parametric gain and bandwidth reported for on-chip parametric amplifiers.

physics.optics↗

arXiv2026

Engineered mode coupling in high-Q microresonators enables deterministic low-repetition-rate soliton microcombs

Soliton optical frequency combs have become key enablers for a wide range of applications, including telecommunications, optical atomic clocks, ultrafast distance measurements, dual-comb spectroscopy, and astrophysical spectrometer calibration, many of which benefit from low repetition rates. However, achieving such low-repetition-rate soliton microcombs is nontrivial as long cavities require substantially higher pump power, which induces stronger thermal effects that, in turn, exacerbate thermal instability and complicate access to stable soliton states. The dual-mode pumping scheme, in which a continuous-wave pump couples to both the comb-generating mode and an auxiliary mode, has proven simple and effective for mitigating thermal instability and enabling thermally accessible soliton generation. Yet, in long-cavity devices, the standard bus-to-resonator coupling conditions for these two modes diverge substantially, resulting in insufficient pump coupling to the auxiliary mode, which makes dual-mode pumping particularly challenging for low-repetition-rate microcombs. In this work, we overcome this limitation by coupling the pump to the auxiliary mode via inter-modal coupling, which can be introduced in racetrack microresonators and engineered by tailoring the cavity bend design. We validate this approach in a high-Q (>$10^7$) silicon nitride microresonator and demonstrate thermally accessible, deterministic single-soliton generation at a repetition rate of 33 GHz. This work provides a simple and robust pathway for generating low-repetition-rate soliton microcombs.

physics.optics↗

arXiv2026

ResPrune: Text-Conditioned Subspace Reconstruction for Visual Token Pruning in Large Vision-Language Models

Large Vision-Language Models (LVLMs) rely on dense visual tokens to capture fine-grained visual information, but processing all these tokens incurs substantial computational and memory overhead during inference. To address this issue, we propose ResPrune, a training-free visual token pruning framework that enables efficient LVLM inference by selecting a compact yet informative subset of visual tokens. ResPrune formulates visual token pruning as a subspace reconstruction problem and employs a greedy subspace expansion strategy guided by residual energy, allowing it to preserve the geometric structure of the original visual token space. To further incorporate cross modal alignment, the selection process is conditioned on textual relevance, encouraging the retention of tokens that are both informative and instruction-relevant. The proposed method is lightweight and model-agnostic, and can be seamlessly integrated into existing LVLM pipelines without retraining or architectural modifications. Extensive experiments on multiple LVLM backbones, including LLaVA-1.5, LLaVA-NeXT, and Qwen2.5-VL, demonstrate that ResPrune consistently outperforms existing pruning approaches across a wide range of benchmarks, while achieving effective reductions in computation, memory consumption, and inference latency.

cs.LG↗

arXiv2026

Discovering Decoupled Functional Modules in Large Language Models

Understanding the internal functional organization of Large Language Models (LLMs) is crucial for improving their trustworthiness and performance. However, how LLMs organize different functions into modules remains highly unexplored. To bridge this gap, we formulate a functional module discovery problem and propose an Unsupervised LLM Cross-layer MOdule Discovery (ULCMOD) framework that simultaneously disentangles the large set of neurons in the entire LLM into modules while discovering the topics of input samples related to these modules. Our framework introduces a novel objective function and an efficient Iterative Decoupling (IterD) algorithm. Extensive experiments show that our method discovers high-quality, disentangled modules that capture more meaningful semantic information and achieve superior performance in various downstream tasks. Moreover, our qualitative analysis reveals that the discovered modules show semantic coherence, correspond to interpretable specializations, and a clear spatial and hierarchical organization within the LLM. Our work provides a novel tool for interpreting the functional modules of LLMs, filling a critical blank in LLM's interpretability research.

cs.LG↗

arXiv2026

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend 910C chips connected by a high-bandwidth UB fabric and global shared memory. It serves models including DeepSeek, Kimi, GLM, Qwen, and MiniMax, among others. xDeepServe is built around Transformerless, a disaggregated execution architecture that decomposes transformer inference into modular units -- attention, feedforward, and MoE -- and supports disaggregated Prefill-Decode and MoE-Attention deployments. To enable disaggregation, we develop XCCL, a memory-semantic communication layer providing microsecond-level point-to-point and scalable all-to-all primitives, and we extend FlowServe with decentralized DP groups and techniques to mitigate stragglers and synchronization variance. In a peak decoding configuration, xDeepServe reaches 2400 tokens/s per Ascend 910C chip at ~50ms time-per-output-token (TPOT).

cs.DC↗

arXiv2026

Tracing Missing Baryons in the Cosmic Filaments with tSZ and CMB-Lensing Stacking

We investigate the distribution of missing baryons in the cosmic filaments by stacking $\sim 30,700$ filaments across the northern and southern SDSS sky regions using Planck Compton-$y$ and CMB lensing maps. Filaments are identified using the DisPerSE algorithm applied to the SDSS LOWZ-CMASS galaxy samples, selecting structures with lengths between 30-100 cMpc and redshifts in the range $0.2 < z < 0.6$. Radial profiles are extracted out to 25 cMpc from the filament spines, and galaxy clusters with halo masses above $\sim 3 \times 10^{13} M_\odot$ are masked to reduce contamination. We detect the thermal Sunyaev-Zeldovich (tSZ) signal at $7.82σ$ and the CMB lensing signal at $7.78σ$. The stacked profiles are corrected by a geometric bias correction based on filament inclination with respect to the line-of-sight, and they are portrayed assuming isothermal, cylindrically symmetric models. We explore different gas and matter density distributions, focusing on the $β$-models with $(α,β) = (2,2/3)$ or $(1,1)$. By jointly fitting the Compton-$y$ and $κ$ profiles, we constrain the central electron overdensity and temperature to be $δ= 4.18^{+2.01}_{-1.06}$ and $T_e = 2.74^{+0.65}_{-0.53}\times 10^6 \mathrm{K}$ for the standard $β$-model. These results suggest that filamentary WHIM in our selected long filaments contributes a significant baryon fraction of $0.127^{+0.019}_{-0.021}\times Ω_b$ to the cosmic baryon budget.

astro-ph.CO↗

arXiv2026

Passive Daytime Radiative Cooling Enabled by Bio-Derived Ceramic-Polymer Coatings on Rapid-Curing Fiberglass Casts

Passive daytime radiative cooling (PDRC) provides an energy-free approach to suppress surface temperatures by reflecting solar irradiation while emitting thermal radiation through the mid-infrared atmospheric window. Despite rapid progress in optical performance, most PDRC systems remain limited by rigid, fragile, or planar substrates, restricting their use on flexible, curved, or wearable surfaces. Here, we report a biocompatible and structurally robust PDRC system integrated onto a commercial rapid-curing fiberglass cast, a conformal substrate widely used in orthopedic and industrial applications. The cooling architecture adopts a bilayer polymer design consisting of a polyvinyl alcohol (PVA) adhesion layer and a polymethyl methacrylate (PMMA) protective layer, both embedded with calcium pyrophosphate (CPP) ceramic particles derived from processed animal bone waste. The bio-derived CPP simultaneously enables broadband solar scattering and high mid-infrared emittance, while offering sustainability and biocompatibility advantages. The resulting composite exhibits over 90% solar reflectance and achieves up to 15 C sub-ambient cooling under direct outdoor sunlight.

physics.optics↗

arXiv2026

3D-Printed Hybrid Liquid-CPCM Cooling Modules for High-Performance Thermal Management of Lithium-Ion Pouch Cells

Efficient thermal management is critical for ensuring the safety, performance, and durability of lithium ion pouch cells (LIPCs), particularly under high power operating conditions where conventional battery thermal management systems (BTMS) struggle to balance cooling effectiveness, structural simplicity, and weight. Here, we report a lightweight hybrid BTMS that synergistically integrates active liquid cooling with composite phase change material (CPCM) based thermal buffering through a 3D printed hexagonal architecture. The system is fabricated via a two step additive manufacturing process that enables sealed CPCM encapsulation and isolated liquid cooling pathways within a single carbon fiber reinforced nylon module, effectively eliminating leakage risks while allowing precise geometric control. Hexagonally partitioned CPCM cavities maximize the CPCM wall interfacial area and shorten internal conduction paths, accelerating latent heat absorption, while embedded serpentine liquid channels provide continuous convective heat removal and prevent CPCM saturation. A nanocarbon enhanced CPCM is employed to overcome the intrinsic low thermal conductivity of conventional paraffin based materials.

eess.SY↗