Search arXiv⌕ Search

arXiv subjects

Yi Liu

Publications and source records attributed to Yi Liu.

At least 199 records · Page 11Linked to original sources
arXiv2026

VideoReasonBench: Can MLLMs Perform Vision-Centric Complex Video Reasoning?

Recent studies have shown that long chain-of-thought (CoT) reasoning can significantly enhance the performance of large language models (LLMs) on complex tasks. However, this benefit is yet to be demonstrated in the domain of video understanding, since most existing benchmarks lack the reasoning depth required to demonstrate the advantages of extended CoT chains. While recent efforts have proposed benchmarks aimed at video reasoning, the tasks are often knowledge-driven and do not rely heavily on visual content. To bridge this gap, we introduce VideoReasonBench, a benchmark designed to evaluate vision-centric, complex video reasoning. To ensure visual richness and high reasoning complexity, each video in VideoReasonBench depicts a sequence of fine-grained operations on a latent state that is only visible in part of the video. The questions evaluate three escalating levels of video reasoning skills: recalling observed visual information, inferring the content of latent states, and predicting information beyond the video. Under such task setting, models have to precisely recall multiple operations in the video, and perform step-by-step reasoning to get correct final answers for these questions. Using VideoReasonBench, we comprehensively evaluate 18 state-of-the-art multimodal LLMs (MLLMs), finding that most perform poorly on complex video reasoning -- e.g., GPT-4o achieves only 6.9% accuracy -- while the thinking-enhanced Gemini-2.5-Pro significantly outperforms others with 56.0% accuracy. Our investigations on "test-time scaling" further reveal that extended thinking budget, while offering none or minimal benefits on existing video benchmarks, is essential for improving the performance on VideoReasonBench.

cs.CV↗
arXiv2026

GeoZero: Incentivizing Reasoning from Scratch on Geospatial Scenes

Multimodal large language models (MLLMs) have undergone rapid development in advancing geospatial scene understanding. Recent studies have sought to enhance the reasoning capabilities of remote sensing MLLMs, typically through cold-start training with elaborately curated chain-of-thought (CoT) data. However, this approach not only incurs substantial annotation costs but also introduces human biases that may limit the diversity of model reasoning. To address these challenges, we propose GeoZero, a framework that enables MLLMs to perform geospatial reasoning without any predefined CoT supervision. Specifically, we construct two datasets, GeoZero-Instruct and GeoZero-Hard. GeoZero-Instruct allows the model to acquire preliminary geospatial knowledge through supervised fine-tuning, while GeoZero-Hard stimulates deep reasoning during the subsequent reinforcement learning stage. Furthermore, we introduce Answer-Anchored Group Relative Policy Optimization (A$^2$GRPO), where the reasoning process is regularized by the model's own answers, encouraging diverse yet accurate thinking. Extensive experiments on multiple remote sensing vision-language benchmarks demonstrate that GeoZero not only surpasses existing state-of-the-art methods but also fosters universal emergent reasoning capabilities across diverse geospatial tasks. Code, data, and models are available at https://github.com/MiliLab/GeoZero.

cs.CV↗
arXiv2026

d-HNSW: A High-performance Vector Search Engine on Disaggregated Memory

Efficient vector search is essential for powering large-scale AI applications, such as LLMs. Existing solutions are designed for monolithic architectures where compute and memory are tightly coupled. Recently, disaggregated architecture breaks this coupling by separating compution and memory resources into independently scalable pools to improve utilization. However, applying vector database on disaggregated memory system brings unique challenges to system design due to its graph-based index. We present d-HNSW, the first RDMA-based vector search engine optimized for disaggregated memory systems. d-HNSW preserves HNSW's high accuracy while addressing the new system-level challenges introduced by disaggregation: 1) network inefficiency from pointer-chasing traversals, 2) non-contiguous remote memory layout induced by dynamic insertions, 3) redundant data transfers in batch workloads, and 4) resource underutilization due to sequential execution. d-HNSW tackles these challenges through a set of hardware-algorithm co-designed techniques, including 1) balanced clustering with a lightweight representative index to reduce network round-trips and ensure predictable latency, 2) an RDMA-friendly graph layout that preserves data contiguity under dynamic insertions, 3) query-aware data loading to eliminate redundant fetches across batch queries, and 4) a pipelined execution model that overlaps RDMA transfers with computation to hide network latency and improve throughput. Our evaluation results in a public cloud show that d-HNSW achieves up to < 10-2x query latency and > 100x query throughput compared to other baselines, while maintaining a high recall of 94%.

cs.DB↗
arXiv2026

Recover to Predict: Progressive Retrospective Learning for Variable-Length Trajectory Prediction

Trajectory prediction is critical for autonomous driving, enabling safe and efficient planning in dense, dynamic traffic. Most existing methods optimize prediction accuracy under fixed-length observations. However, real-world driving often yields variable-length, incomplete observations, posing a challenge to these methods. A common strategy is to directly map features from incomplete observations to those from complete ones. This one-shot mapping, however, struggles to learn accurate representations for short trajectories due to significant information gaps. To address this issue, we propose a Progressive Retrospective Framework (PRF), which gradually aligns features from incomplete observations with those from complete ones via a cascade of retrospective units. Each unit consists of a Retrospective Distillation Module (RDM) and a Retrospective Prediction Module (RPM), where RDM distills features and RPM recovers previous timesteps using the distilled features. Moreover, we propose a Rolling-Start Training Strategy (RSTS) that enhances data efficiency during PRF training. PRF is plug-and-play with existing methods. Extensive experiments on datasets Argoverse 2 and Argoverse 1 demonstrate the effectiveness of PRF. Code is available at https://github.com/zhouhao94/PRF.

cs.RO↗
arXiv2026

Soft Phonon Charge-Density Wave Formation in the Kagome Metal KV$_3$Sb$_5$

A range of of unusual emergent behaviors have been reported in the charge-density wave (CDW) state of the $A$V$_3$Sb$_5$ ($A=~$K, Rb, Cs) kagome metals, including a CDW formation process without soft phonons, which points to an unconventional CDW mechanism. Here, we use inelastic x-ray scattering to show that the CDW in KV$_3$Sb$_5$ forms via phonons that soften to zero energy at the CDW ordering vector ($L$-point) around $T_{\rm CDW}=78$~K. The intensity of soft phonons exhibit a remarkable in-plane anisotropy, extending over a much larger momentum range along $L$-$A$ relative to $L$-$H$, which leads to diffuse scattering common among $A$V$_3$Sb$_5$. Using first-principles calculations, we find that the momentum-dependent electron-phonon coupling (EPC) is peaked at $L$ and exhibits the same in-plane anisotropy as the phonon softening. Conversely, the electronic susceptibility is not peaked at $L$ and shows the opposite in-plane anisotropy. Our findings favor momentum-dependent EPC as the driving mechanism of the CDW in KV$_3$Sb$_5$, with a CDW formation process similar to that of transition metal dichalcogenides.

cond-mat.supr-con↗
arXiv2026

TaPD: Temporal-adaptive Progressive Distillation for Observation-Adaptive Trajectory Forecasting in Autonomous Driving

Trajectory prediction is essential for autonomous driving, enabling vehicles to anticipate the motion of surrounding agents to support safe planning. However, most existing predictors assume fixed-length histories and suffer substantial performance degradation when observations are variable or extremely short in real-world settings (e.g., due to occlusion or a limited sensing range). We propose TaPD (Temporal-adaptive Progressive Distillation), a unified plug-and-play framework for observation-adaptive trajectory forecasting under variable history lengths. TaPD comprises two cooperative modules: an Observation-Adaptive Forecaster (OAF) for future prediction and a Temporal Backfilling Module (TBM) for explicit reconstruction of the past. OAF is built on progressive knowledge distillation (PKD), which transfers motion pattern knowledge from long-horizon "teachers" to short-horizon "students" via hierarchical feature regression, enabling short observations to recover richer motion context. We further introduce a cosine-annealed distillation weighting scheme to balance forecasting supervision and feature alignment, improving optimization stability and cross-length consistency. For extremely short histories where implicit alignment is insufficient, TBM backfills missing historical segments conditioned on scene evolution, producing context-rich trajectories that strengthen PKD and thereby improve OAF. We employ a decoupled pretrain-reconstruct-finetune protocol to preserve real-motion priors while adapting to backfilled inputs. Extensive experiments on Argoverse 1 and Argoverse 2 show that TaPD consistently outperforms strong baselines across all observation lengths, delivers especially large gains under very short inputs, and improves other predictors (e.g., HiVT) in a plug-and-play manner. Code will be available at https://github.com/zhouhao94/TaPD.

cs.CV↗
arXiv2026

Time-reversal symmetry breaking superconductivity with electronic glass in nickelate (La, Pr, Sm)3Ni2O7 films

The discovery of Ruddlesden-Popper (R-P) nickelate superconductors under high pressure heralds a new chapter of high-transition temperature (high-Tc) superconductivity. Recently, ambient-pressure superconductivity is achieved in R-P bilayer nickelate thin films through epitaxial compressive strain, unlocking the potential for understanding the nature of the unconventional superconductivity. Here, through electrical transport study, we report the discovery of time-reversal symmetry (TRS) breaking superconductivity with electronic glass in bilayer nickelate (La, Pr, Sm)3Ni2O7 films. It emerges in the lower-temperature regime of superconducting transition to the zero-resistance state, and is captured by three remarkable characteristics: 1. Unconventional magnetoresistance hysteresis, the direct evidence of TRS breaking, which is robust under different magnetic field orientations and differs fundamentally from trapped vortices or long-range-ordered magnetism. Successive oxygen reductions simultaneously weaken both the superconductivity and hysteresis, revealing their mutual connections to selective electronic orbitals. 2. Magnetic field history-dependence and zero-field non-reciprocity in the current-voltage responses, further substantiating the intrinsic and spontaneous TRS breaking. 3. Logarithmically slow resistance relaxations upon the removal of magnetic field, the hallmarks of glassy dynamics. Distinguished by the striking magnetic field history- and time-dependent properties, our findings uncover an unprecedented superconducting state in the nickelate superconductors, providing phenomenological and conceptual advances for future research on high-Tc superconductivity.

cond-mat.supr-con↗
arXiv2026

Seeing Clearly without Training: Mitigating Hallucinations in Multimodal LLMs for Remote Sensing

Multimodal large language models (MLLMs) suffer from pronounced hallucinations in remote sensing visual question-answering (RS-VQA), primarily caused by visual grounding failures in large-scale scenes or misinterpretation of fine-grained small targets. To systematically analyze these issues, we introduce RSHBench, a protocol-based benchmark for fine-grained diagnosis of factual and logical hallucinations. To mitigate grounding-induced factual hallucinations, we further propose Relative Attention-Driven Actively Reasoning (RADAR), a training-free inference method that leverages intrinsic attention in MLLMs to guide progressive localization and fine-grained local reasoning at test time. Extensive experiments across diverse MLLMs demonstrate that RADAR consistently improves RS-VQA performance and reduces both factual and logical hallucinations. Code and data will be publicly available at: https://github.com/MiliLab/RADAR

cs.CV↗
arXiv2026

In-Orbit GRB Identification Using LLM-based model for the CXPD CubeSat

To validate key technologies for wide field-of-view (FOV) X-ray polarization measurements, the Cosmic X-ray Polarization Detector (CXPD) CubeSat series has been developed as a prototype platform for the Low-Energy X-ray Polarization Detector (LPD) onboard the POLAR-2 mission. The wide-FOV design significantly increases the complexity of the background environment, posing notable challenges for real-time gamma-ray burst (GRB) identification. In this work, we propose an in-orbit GRB identification method based on machine learning, using simulated spectral data as input. A training dataset was constructed using a Geant4-based simulator, incorporating in-orbit background and GRB events modeled within the 2-10 keV energy range. To meet the computational constraints of onboard processing, we employ a multimodal large language model (MLLM), which is fine-tuned using low-rank adaptation (LoRA) based on miniCPM-V2.6 and quantized to 4-bit precision. The model achieves perfect classification accuracy on validation data and demonstrates strong regression performance in estimating GRB spectral indices, with an RMSE of 0.118. Furthermore, we validate the feasibility of onboard deployment through a simulated satellite data processing pipeline, highlighting the potential of our approach to enable future real-time GRB detection and spectral analysis in orbit.

astro-ph.IM↗
arXiv2026

Data Selection for LLM Alignment Using Fine-Grained Preferences

Large language models (LLMs) alignment aims to ensure that the behavior of LLMs meets human preferences. While collecting data from multiple fine-grained, aspect-specific preferences becomes more and more feasible, existing alignment methods typically work on a single preference and thus struggle with conflicts inherent in such aggregated datasets. As one early attempt, in this paper, we propose a data-centric approach to align LLMs through the effective use of fine-grained preferences. Specifically, we formulate the problem as a direct fine-grained preference optimization and introduce preference divergence (PD) that quantifies inter-aspect preference conflicts. Instead of directly tackling the consequent complicated optimization, we recast it as a data selection problem and propose a simple yet effective strategy, which identifies a subset of data corresponding to the most negative PD values, for efficient training. We theoretically analyze the loss-bound optimality of our selection strategy and conduct extensive empirical studies on varied settings and datasets to demonstrate that our practical selection method could achieve consistent improvement against standard full-data alignment, using even just 30% of the data. Our work shares a line that LLM alignment using fine-grained preferences is highly feasible.

cs.LG↗
arXiv2026

Emergent quantum phenomena via phase-coherence engineering in infinite-layer nickelate superconductors

Dimensionality of a physical system, conventionally an invariant geometric characteristic, fundamentally governs the universality class of phase transitions and the landscape of emergent collective phenomena. In low-dimensional or layered high-temperature superconductors, the macroscopic phase coherence of superconducting orders is typically confined in two dimensions, underscoring the critical role of phase fluctuations in determining the overall phase diagrams. Here, we strategically enhance the phase fluctuations by fabricating periodically arranged nano-holes in the infinite-layer nickelate superconducting films, effectively constructing Josephson junction arrays. In the nano-patterned films, the weakening of macroscopic phase coherence drives a two-stage superconducting transition towards an anomalous metallic ground state with saturated resistance. The emergence of charge-2e quantum oscillations manifests the coherence across the array, while an anomalous zero-field magnetoresistance peak signifies the extreme quantum phase fluctuations persisting to ultralow temperatures. Remarkably, with quantum fluctuations enhanced synergistically by nano-patterning and magnetic fields, an anomalous reversal of superconducting anisotropy is observed in Nd-nickelates, where in-plane critical fields fall below out-of-plane values. The evolution of anisotropy may unmask an internal exchange-Zeeman field coupled to the collective electronic states. Our results unveil how superconductivity evolves in response to phase fluctuations, establishing nano-patterning as a powerful paradigm to uncover hidden intertwined orders in strongly correlated systems.

cond-mat.supr-con↗
arXiv2026

Enhancing Vision-Language Navigation with Multimodal Event Knowledge from Real-World Indoor Tour Videos

Vision-Language Navigation (VLN) agents often struggle with long-horizon reasoning in unseen environments, particularly when facing ambiguous, coarse-grained instructions. While recent advances use knowledge graph to enhance reasoning, the potential of multimodal event knowledge inspired by human episodic memory remains underexplored. In this work, we propose an event-centric knowledge enhancement strategy for automated process knowledge mining and feature fusion to solve coarse-grained instruction and long-horizon reasoning in VLN task. First, we construct YE-KG, the first large-scale multimodal spatiotemporal knowledge graph, with over 86k nodes and 83k edges, derived from real-world indoor videos. By leveraging multimodal large language models (i.e., LLaVa, GPT4), we extract unstructured video streams into structured semantic-action-effect events to serve as explicit episodic memory. Second, we introduce STE-VLN, which integrates the above graph into VLN models via a Coarse-to-Fine Hierarchical Retrieval mechanism. This allows agents to retrieve causal event sequences and dynamically fuse them with egocentric visual observations. Experiments on REVERIE, R2R, and R2R-CE benchmarks demonstrate the efficiency of our event-centric strategy, outperforming state-of-the-art approaches across diverse action spaces. Our data and code are available on the project website https://sites.google.com/view/y-event-kg/.

cs.RO↗
arXiv2026

GAGA: Gaussianity-Aware Gaussian Approximation for Efficient 3D Molecular Generation

Gaussian Probability Path based Generative Models (GPPGMs) generate data by reversing a stochastic process that progressively corrupts samples with Gaussian noise. Despite state-of-the-art results in 3D molecular generation, their deployment is hindered by the high cost of long generative trajectories, often requiring hundreds to thousands of steps during training and sampling. In this work, we propose a principled method, named GAGA, to improve generation efficiency without sacrificing training granularity or inference fidelity of GPPGMs. Our key insight is that different data modalities obtain sufficient Gaussianity at markedly different steps during the forward process. Based on this observation, we analytically identify a characteristic step at which molecular data attains sufficient Gaussianity, after which the trajectory can be replaced by a closed-form Gaussian approximation. Unlike existing accelerators that coarsen or reformulate trajectories, our approach preserves full-resolution learning dynamics while avoiding redundant transport through truncated distributional states. Experiments on 3D molecular generation benchmarks demonstrate that our GAGA achieves substantial improvement on both generation quality and computational efficiency.

cs.LG↗
arXiv2026

Momentum Measurement of Charged Particles in FASER's Emulsion Detector at the LHC

We present a momentum measurement method based on multiple Coulomb scattering (MCS) in the FASER$ν$ emulsion detector. The measurement of charged-particle momenta is essential for studying neutrino interactions in the TeV energy range at the FASER experiment. This method exploits the sub-micron spatial resolution and long tracking length of the FASER$ν$ detector, enabling momentum determination from a few GeV up to a few TeV. The performance was evaluated using Geant4-based Monte Carlo simulations and validated with muon test beam data in the momentum range 100-300 GeV. As a first probe of the method for higher momentum muons, background muons recorded by the FASER$ν$ detector were examined, showing reconstructed momenta consistent with expectations from their angular spread.

hep-ex↗
arXiv2026

What Makes a Good LLM Agent for Real-world Penetration Testing?

LLM-based agents show promise for automating penetration testing, yet reported performance varies widely across systems and benchmarks. We analyze 28 LLM-based penetration testing systems and evaluate five representative implementations across three benchmarks of increasing complexity. Our analysis reveals two distinct failure modes: Type A failures stem from capability gaps (missing tools, inadequate prompts) that engineering readily addresses, while Type B failures persist regardless of tooling due to planning and state management limitations. We show that Type B failures share a root cause that is largely invariant to the underlying LLM: agents lack real-time task difficulty estimation. As a result, agents misallocate effort, over-commit to low-value branches, and exhaust context before completing attack chains. Based on this insight, we present Excalibur, a penetration testing agent that couples strong tooling with difficulty-aware planning. A Tool and Skill Layer eliminates Type A failures through typed interfaces and retrieval-augmented knowledge. A Task Difficulty Assessment (TDA) mechanism addresses Type B failures by estimating tractability through four measurable dimensions (horizon estimation, evidence confidence, context load, and historical success) and uses these estimates to guide exploration-exploitation decisions within an Evidence-Guided Attack Tree Search (EGATS) framework. Excalibur achieves up to 91% task completion on CTF benchmarks with frontier models (39 to 49% relative improvement over baselines) and compromises 4 of 5 hosts on the GOAD Active Directory environment versus 2 by prior systems. These results show that difficulty-aware planning yields consistent end-to-end gains across models and addresses a limitation that model scaling alone does not eliminate.

cs.CR↗
arXiv2026

Revisiting self-seeding mechanism by generating vector ultraviolet N$_2^{+}$ lasing

An intense femtosecond laser pulse can generate ultraviolet air lasing, offering a promising remote light source. A long-standing hypothesis is whether it is seeded by a self-generated spectral component, such as the second harmonic that is inevitably produced by the plasma gradient. Here, we report the generation of both radially and azimuthally polarized N$_2^+$ lasing driven by a single 800-nm cylindrical vector beam. Meanwhile, the same vector pump was applied to drive the generation of vector second harmonics in plasma. The radially polarized pump produces radially polarized second harmonics while the azimuthally polarized pump yields no second harmonic generation owing to the radial direction of plasma gradient. The absence of the azimuthally polarized second harmonic rules out the hypothesis of self-seeding by second harmonics, as both radially and azimuthally polarized N$_2^+$ lasing are observed with comparable intensities. By characterizing the spatial phase distribution of vector 391-nm lasing, we concluded that its phase is synchronized with the pump. These results suggest that amplified spontaneous emissions are the origin of N$_2^+$ lasing under the most common condition of low gas pressure, which was effectively demonstrated by theoretical simulations. Our work provides a promising method for remotely generating vector ultraviolet light sources.

physics.atom-ph↗
arXiv2026

Estimation and Inference of the Win Ratio for Two Hierarchical Endpoints Subject to Censoring and Missing Data

The win ratio (WR) is a widely used metric to compare treatments in randomized clinical trials with hierarchically ordered endpoints. Counting-based approaches, such as Pocock's algorithm, are the standard for WR estimation. However, this algorithm treats participants with censored or missing data inadequately, which may lead to biased and inefficient estimates, particularly in the presence of heterogeneous censoring or missing data between treatment groups. Although recent extensions have addressed some of these limitations for hierarchical time-to-event endpoints, no existing methods -- aside from the computationally intensive multiple imputation approach -- can accommodate settings that include non-survival endpoints that are subject to missing data. In this paper, we propose a simple nonparametric maximum likelihood estimator (NPMLE) of WR for two hierarchical endpoints that are subject to censoring and missing data. Our method uses all observed data, avoids strong parametric assumptions, and comes with a closed-form asymptotic variance estimator. We demonstrate its performance using simulation studies and two data examples, based on the HEART-FID and ISCHEMIA trials. The proposed method provides a consistent estimator, improves estimation efficiency, and is robust under non-informative censoring and missing at random (MAR) assumptions, offering a flexible alternative to existing WR estimation methods. A user-friendly R package, WinRS, is available to facilitate implementation.

stat.ME↗
arXiv2026

PlanetServe: A Decentralized, Scalable, and Privacy-Preserving Overlay for Democratizing Large Language Model Serving

While significant progress has been made in research and development on open-source and cost-efficient large-language models (LLMs), serving scalability remains a critical challenge, particularly for small organizations and individuals seeking to deploy and test their LLM innovations. Inspired by peer-to-peer networks that leverage decentralized overlay nodes to increase throughput and availability, we propose GenTorrent, an LLM serving overlay that harnesses computing resources from decentralized contributors. We identify four key research problems inherent to enabling such a decentralized infrastructure: 1) overlay network organization; 2) LLM communication privacy; 3) overlay forwarding for resource efficiency; and 4) verification of serving quality. This work presents the first systematic study of these fundamental problems in the context of decentralized LLM serving. Evaluation results from a prototype implemented on a set of decentralized nodes demonstrate that GenTorrent achieves a latency reduction of over 50% compared to the baseline design without overlay forwarding. Furthermore, the security features introduce minimal overhead to serving latency and throughput. We believe this work pioneers a new direction for democratizing and scaling future AI serving capabilities.

cs.DC↗