Searcharxiv⌕ Search

arXiv subjects

Jie Liu

Publications and source records attributed to Jie Liu.

At least 37 records · Page 2Linked to original sources
arXiv2026

Found but Not Read: When Extracted Text Closes the Retrieval-Reading Gap in Document Vision-Language Models

Retrieval-augmented document question answering assumes that once the right page is found, a vision-language model (VLM) can read it. We show that this assumption often fails, leaving a retrieval-reading gap: evidence found but not used. A paired protocol isolates this gap by comparing answers from the retrieved page images alone with answers from the same images plus their extracted text. On FoveDoc-Bench, our benchmark with traceable evidence, retrieval finds nearly every evidence page, yet adding CPU-OCR text raises strict accuracy by 13 to 16 points. An exact text layer roughly doubles the gain, which appears across six VLMs from three families and, within one family, narrows with scale without closing. The reader can read this evidence but cannot find it: crops of it recover most of the text gain, boxes around it on the page none. The same protocol identifies two boundaries. Extracted text helps on textual evidence but is neutral or harmful on charts and figures. Its advantage shrinks as retrieval degrades, and unrelated text of the same form adds nothing detectable. Extracted text is an amplifier of retrieval that works, not a substitute for retrieval that does not. Our code is available at https://github.com/atoz03/fovedoc-sup.

cs.CV↗
arXiv2026

Guide, Think, Act: Interactive Embodied Reasoning in Vision-Language-Action Models

In this paper, we propose GTA-VLA(Guide, Think, Act), an interactive Vision-Language-Action (VLA) framework that enables spatially steerable embodied reasoning by allowing users to guide robot policies with explicit visual cues. Existing VLA models learn a direct "Sense-to-Act" mapping from multimodal observations to robot actions. While effective within the training distribution, such tightly coupled policies are brittle under out-of-domain (OOD) shifts and difficult to correct when failures occur. Although recent embodied Chain-of-Thought (CoT) approaches expose intermediate reasoning, they still lack a mechanism for incorporating human spatial guidance, limiting their ability to resolve visual ambiguities or recover from mistakes. To address this gap, our framework allows users to optionally guide the policy with spatial priors, such as affordance points, boxes, and traces, which the subsequent reasoning process can directly condition on. Based on these inputs, the model generates a unified spatial-visual Chain-of-Thought that integrates external guidance with internal task planning, aligning human visual intent with autonomous decision-making. For practical deployment, we further couple the reasoning module with a lightweight reactive action head for efficient action execution. Extensive experiments demonstrate the effectiveness of our approach. On the in-domain SimplerEnv WidowX benchmark, our framework achieves a state-of-the-art 81.2% success rate. Under OOD visual shifts and spatial ambiguities, a single visual interaction substantially improves task success over existing methods, highlighting the value of interactive reasoning for failure recovery in embodied control. More details of the project can be found here: https://github.com/FutianLabs/GTA-VLA.

cs.RO↗
arXiv2026

CoFiE: Coarse-to-Fine Evidence Selection for Efficient Streaming Video Understanding

Streaming video understanding requires Vision Language Models (VLLMs) to process growing video streams and answer user questions under tight latency constraints. Existing methods improve efficiency through token pruning and memory-bank schemes, but mainly reduce visual tokens after visual encoding. Consequently, downstream token pruning alone cannot substantially reduce end-to-end latency because the expensive frame encoding cost has already been incurred. We propose CoFiE, a Coarse-to-Fine Evidence Selection framework that decouples evidence selection into a coarse, query-agnostic filtering stage before the vision encoder and a fine, query-specific refinement stage during LLM prefill. CoFiE introduces Novelty-Guided Frame Filtering to retain visually distinctive candidate frames and Query-Specific Evidence Refinement to select the frames most relevant to the user query. This design removes substantial redundancy before frame encoding while preserving query-specific refinement once semantic information becomes available. Experiments show that CoFiE establishes a new state-of-the-art accuracy-efficiency trade-off across multiple video understanding benchmarks, reaching 78.86% accuracy on StreamingBench and 68.72% on OvO-Bench, with improvements of up to 3.15% over prior methods. Even with up to 80% evidence-frame filtering, CoFiE outperforms strong open-source multimodal models while improving end-to-end inference latency by up to 2.54 times.

cs.CV↗
arXiv2026

NVV-Locator: From Transcript Tags to Acoustic Boundaries for Fine-Grained Nonverbal Vocalization Grounding

Human speech includes nonverbal vocalizations (NVVs), such as laughter, sighs, breaths, and coughs, which convey affective and interactional information. Existing approaches typically represent NVVs as transcript-level tags, providing limited supervision for their waveform-time boundaries. We present NVV-Locator for fine-grained NVV temporal grounding. We first unify 26 NVV categories across public resources and construct large-scale timestamp-supervised training data through dual-LLM verification, transcript-guided forced alignment, and energy-based boundary refinement. We further introduce NVV-TimeBench, an expert-refined benchmark with 667 utterances and 1,094 events. NVV-Locator uses a non-autoregressive slot-filling architecture to jointly predict lexical timestamps, NVV categories, and event boundaries. On NVV-TimeBench, it achieves 71.0% Micro F1, 70.2% Macro F1, 80.4% Macro mIoU, and 59.6 ms Macro mMAE, outperforming the evaluated large audio model counterparts. Evaluation on an external corpus further demonstrates the cross-corpus generalization of NVV-Locator.

eess.AS↗
arXiv2026

VERA: Verifiable Feasibility Representations with Counterfactual Credit for Constrained Multi-Agent Control

Constrained multi-agent control requires more than predicting rewarding actions: an action can cease to be executable as contact windows, shared capacity, and deadlines change. We introduce VERA, a centralized-training, decentralized-execution framework that separates feasibility estimation from credit assignment. Each actor predicts a five-dimensional verifiable feasibility representation (VFR). After an action is proposed, exact action-conditioned margins available only during training supervise that representation, while a counterfactual group-relative advantage (CGRA) ranks candidate representation-action pairs. Execution uses one actor pass and no privileged state. In a dynamic space-air-ground integrated network (SAGIN), VERA obtains 55.33% +/- 3.60% success with 0.45% +/- 0.81% coverage violation, within 1.33 percentage points of a privileged-mask reference. With rewards matched over ten paired seeds, VERA improves success over the strongest baseline by 8.74 percentage points (p=0.023) and reduces violation by 52.19 percentage points (p=5.7e-8). A ten-seed 4-by-2 factorial attributes a 14.16-16.48 percentage-point gain to CGRA across handcrafted, learned, random, and latent representations; evaluation on seven unseen topologies preserves a 24.33-30.02 percentage-point advantage over multi-agent proximal policy optimization. From 10 to 40 users, success remains 50.1-53.8%, and VFR adds only 0.026 ms to a central processing unit (CPU) actor step. Cross-domain tests further identify the governing condition: counterfactual credit succeeds when candidate scores respect shared constraints and fails under incompatible reward geometries. These results establish action-conditioned feasibility as an auditable training interface and counterfactual credit as a geometry-dependent optimization mechanism.

cs.LG↗
arXiv2026

What Can a Leaderboard Certify? Compositional Controllability for Fair Evaluation and Training of Biomedical Literature-Review Agents

Leaderboards rank long-horizon agents by their final outputs. Yet a higher score alone does not establish whether two systems are comparable or which stage accounts for the difference. Unequal evidence, inputs, or budgets can affect scores, and statistical corrections do not remove this mismatch. We introduce compositional controllability to address these questions. A comparison window covers one stage, several stages, or the whole agent. Our central result bounds the gap between observed and controlled score differences using only nuisance outside the window. This yields an admissibility test applied before scores are inspected. Inadmissible comparisons are refused. For admissible pairs, an ordering is certified only when the score gap exceeds the combined sampling and nuisance radii; otherwise, it remains undecided. These decisions give each system a rank interval. We introduce BioLitBench, a benchmark of 2,042 biomedical articles represented as structured claim graphs. Among seven published pipelines, a conventional statistical analysis declares a winner in 14 of 21 pairwise comparisons. Yet the top-ranked system alone received the target review's bibliography. To isolate pipeline performance, our test requires matched inputs and a fixed backbone model. It refuses 11 of the 21 comparisons, including every comparison involving the top-ranked system. Seven of the 14 conventional conclusions fall within these refused pairs. The same comparison windows support stage-level training. We train SCRIBE on Qwen3.8-27B using rewards measured at each stage's exit. Under matched evidence, SCRIBE achieves a certified rank interval of [1,2], with certified advantages over all evaluated published pipelines and the evaluated Claude and OpenAI agents. Under same pool, SCRIBE matches the strongest published retriever and is certified above three published pipelines.

cs.LG↗
arXiv2026

IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis

Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for identifying information needs and a Synthesizer for integrating evidence into an evolving summary state. This design separates planning from synthesis while using the summary as the persistent state of search, reducing both capability coupling and context noise. To train IterSynth effectively, we further introduce Role-Decoupled Policy Optimization (RDPO) for reinforcement learning, which combines terminal outcome rewards with turn-level rubric evaluations and computes role-specific advantages for more precise credit assignment. Experiments on five long-horizon deep-search benchmarks such as BrowseComp and Xbench-DS show that IterSynth-8B achieves an average score of 50.7, surpassing the strongest prior $\leq$8B agent by +4.2\%. Moreover, IterSynth serves as a model-agnostic prompting paradigm, delivering substantial zero-shot gains over ReAct and similar prompting paradigms on frontier proprietary models.

cs.CL↗
arXiv2026

Subnanometer thermodynamic overlayers on bimetallic nanoparticles

Nanoparticle properties are governed by their surfaces, yet their atomic-scale surface structures remain challenging to predict. Here, we report that bimetallic nanoparticles can form a thermodynamically controlled "shell-dimer" architecture in which one metal forms an overlayer only a few atomic layers thick on another. Using Au-Rh as a model system, atomic-resolution imaging and molecular dynamics show that an ultrathin Au overlayer forms on Rh, stabilized by competition among surface, interfacial, and strain energies. Anisotropic strain limits its growth to the subnanometer scale, and changes in surface chemistry can destabilize the overlayer altogether. Across a range of bimetallic nanoparticles, we found that overlayer formation is associated with elemental immiscibility and lattice mismatch. These findings provide a basis for understanding and controlling surface structures in multimetallic nanoparticles.

cond-mat.mtrl-sci↗
arXiv2026

Nonlinear polarization effects on plasma screening for fusion reactions

Using a finite-temperature two-center Thomas-Fermi-Dirac model, we investigate nonlinear plasma screening effects on the fusion reactions D-T, $^{12}$C-$^{12}$C, and p-$^{11}$B. The effective interaction between a reacting ion pair is obtained self-consistently at the mean-field level. Combining this potential with a complex Woods-Saxon nuclear potential, we solve the stationary Schrödinger equation to evaluate tunneling probabilities and reaction rates. Compared with Debye-Hückel theory, the two-center screening potential can be stronger or weaker, and the resulting fusion enhancement correspondingly amplified or suppressed, depending on the plasma conditions and the internuclear separation. We identify the underlying mechanisms: relative to the single-center prediction, the nonlinear evacuation of background ions suppresses screening, whereas nonlinear electron accumulation enhances it. The two channels respond to different plasma state parameters: the ionic suppression deepens with increasing coupling strength, whereas the electronic enhancement peaks at intermediate degeneracy. The two-center correction therefore follows a factorized scaling law in the nuclear charges of the reacting pair and the plasma state parameters. These findings highlight the importance of a self-consistent two-center treatment for determining effective interactions and evaluating static screening corrections to fusion reaction rates.

physics.plasm-ph↗
arXiv2026

Harmonic Vector Fields and Betti Numbers in Bounded Three-Dimensional Electromagnetic Domains

The topology of a bounded three-dimensional domain can strongly affect electromagnetic fields. In a topologically complex domain, a curl-free field may not have a globally single-valued scalar potential and a divergence-free field may also fail to have a global vector potential. These topological effects lead naturally to harmonic vector fields in the Helmholtz decomposition. This paper gives a geometric and constructive study of such fields using vector analysis, circulation integrals, cutting surfaces, scalar Laplace problems, Stokes' theorem, and Green's identity. The first Betti number is interpreted through independent handle-type circulations, whereas the second Betti number counts enclosed voids. Direct proofs are given for the dimensions of the Neumann and Dirichlet harmonic-field spaces. The analysis is also extended to anisotropic lossless media. The results provide a simple topological interpretation of harmonic fields and physical DC modes in bounded electromagnetic resonators.

physics.class-ph↗
arXiv2026

Local obstructions to the surjectivity of spectral morphisms

Chen and Ngo showed that the Hitchin morphism for Higgs bundles on a projective manifold $X$ factors through a closed subscheme of the Hitchin base called the spectral base, and the induced map is called the spectral morphism. In this paper, we investigate local Cohen-Macaulay obstructions to the surjectivity of spectral morphisms via a refined spectral correspondence. We show that every affine normal variety with a local ring that admits no rank one maximal Cohen-Macaulay modules can be geometrically realized as a local obstruction to the surjectivity of spectral morphisms, and that these local obstructions can be compactified via the Kawamata covering trick. As an application of these constructions, we provide counterexamples to the Chen-Ngo conjecture on the surjectivity of spectral morphisms for $GL_r$-Higgs bundles in dimensions $n\geq 9$. Moreover, we construct counterexamples to the Chen-Ngo conjecture for $SL_2$-Higgs bundles in every dimension $n\geq 2$ and for semistable $GL_3$-Higgs bundles on a smooth projective surface.

math.AG↗
arXiv2026

Pre-PEFT Probing: Weight Statistics and Perturbation Robustness for Layer Selection in VLM Vision Encoders

We propose a pre-fine-tuning probing method for Parameter-Efficient Fine-Tuning (PEFT) layer selection, aiming to obtain more stable and higher gains with fewer trainable parameters when adapting large vision--language models (VLMs). Unlike the common practice of applying LoRA and other adapters to all layers at once---where layer selection often relies on heuristic rules---we focus on the vision encoder and directly evaluate the "adaptability'' of each Transformer layer. Specifically, we characterize each layer from two perspectives: (i) the statistical properties of its Q/K/V projection weights (e.g., norms and condition numbers); (ii) robustness under controlled parameter perturbations. We then systematically compare these indicators with the downstream performance gains brought by applying PEFT to a single layer. Across experiments covering seven benchmarks and five PEFT variants, we observe a consistent correlation: layers (or matrices) with larger weight norms and higher condition numbers are usually more robust to perturbations and are more likely to yield larger fine-tuning gains. These results show that distribution-statistics analysis and perturbation tests before fine-tuning can provide practical signals for adaptation-layer selection, thereby maintaining or improving performance while reducing trainable parameters.

cs.CV↗
arXiv2026

Tools-CC-Bench: a Benchmark Suite for Collective Communication with Compression in HPC and AI Workloads

Distributed HPC and LLM workloads increasingly require efficient communication for scalability, yet growing data movement has become a major performance bottleneck. Communication compression can reduce this overhead and complement execution-level optimizations, but its benefits remain difficult to assess because existing benchmarks lack support for diverse backends, realistic datasets, application-specific accuracy metrics, and overlap-induced resource contention. We present CC-Bench, a lightweight, extensible, and application-oriented benchmark suite for evaluating communication compression under realistic execution conditions. CC-Bench uses declarative application-environment modeling to decouple profiling logic from communication libraries, datasets, and fidelity metrics, enabling portable cross-library evaluation. It further combines function-level interception and hardware counter monitoring to characterize per-phase latency, hardware utilization, numerical fidelity, and computation interference. With representative datasets from HPC and LLM workloads, CC-Bench evaluates three compression-enabled communication libraries on CPU and GPU clusters, revealing accuracy-performance trade-offs and bottlenecks to guide practical deployment and optimization.

cs.DC↗
arXiv2026

Elementary modules of the generalized Kronecker quivers K(n)

Let $k$ be an algebraically closed field. The generalized or $n$-Kronecker quiver $K(n)$ is the quiver with two vertices, called a source and a sink, and $n$ arrows from source to sink. We use $\modd\cK_n$ to denote the category of the finite-dimensional modules of the path algebra $\cK_n=kK(n)$. There exist a function called dimension vector $\dimu: \modd \cK_n\rightarrow \mathbb{Z}^2; M \mapsto (\dim_k M_1, \dim_k M_2 )$, and two functors $σ: \mathbb{Z}^2 \rightarrow \mathbb{Z}^2; (x,y)\mapsto (nx-y,x)$, and $δ: \mathbb{Z}^2\rightarrow \mathbb{Z}^2;(x,y)\mapsto (y,x)$. Then the set $\mathbf{F}=\{(x,y)\mid \frac{2}{n}x\leq y\leq x\}$ is the fundamental domain of the dimension vectors of regular modules under the action of $δ$ and $σ$ in $\modd\cK_n$. We call a regular module $M\in \modd\cK_n$ \textit{elementary} if there does not exist a short exact sequence $(0)\rightarrow L\rightarrow M\rightarrow N\rightarrow (0)$ with $L,N$ being non-zero regular. In this note, we focus on the set $\mathbf{F}$, and when $x<n$, we establish a one-to-one correspondence between elementary modules and spaces of fixed rank.

math.RT↗
arXiv2026

Graded Kronecker Modules of Constant Rank

Let $K(n)$ be the generalized Kronecker quiver, and let $(T(n),Ω)$ be the universal covering of $K(n)$, where $n\geq 3$ and $Ω$ is a bipartite orientation. We introduce three different types of representations for quiver $(T(n),Ω)$ by $K(n)$: maps are injective in a representation maps are all surjective in a representation, and representations are of constant rank. We use $\Inj, \Sur$ and $\GR$ to denote the corresponding subcategories containing these representations, respectively, and we show that \begin{center} $\GR= \Inj\cup\Sur$. \end{center}

math.RT↗
arXiv2026

Online Video Agent Harness for Long Video Understanding

Long video understanding often behaves like a visual needle-in-a-haystack problem: query-relevant evidence is sparsely distributed across long temporal spans, while packing dense frames into a single VLM context incurs \textit{context rot} and high cost. Existing video agents often rely on query-agnostic offline preprocessing or ad hoc tool sets, which can miss query-specific details and waste computation. In this work, we present VideoXAgent, a purely online video-agent harness for long video understanding that starts from the given video file and user query, plans and decomposes the task, invokes specialized expert tools on demand, and aggregates multimodal evidence to produce a final answer while resolving conflicts among observations. To support this on-demand invocation, we design a suite of heterogeneous expert tools guided by a data-driven taxonomy of atomic capabilities, spanning scripts, VLMs, and domain models (e.g., detection, OCR, ASR, face recognition). The harness further enforces objective evidence prompting and budget-aware control to curb hallucination and non-termination. Across Video-MME-Long, LongVideoBench-Long, LVBench, and MINERVA, VideoXAgent is competitive with frontier LMMs and video agents under a smaller context footprint---about 50k tokens of agent context per sample, even on hour-long videos. In particular, on complex video-reasoning benchmarks such as MINERVA, it matches this level while using only about 15\% of the context of a 1,024-frame dense-packing baseline. Notably, the harness remains effective with a visually weak or even text-only orchestrator, suggesting that strong long-video understanding can emerge from progressive agentic evidence seeking rather than from packing the full video into a single context. Project page: https://go-agent-x.github.io/video_agent_harness/

cs.CV↗
arXiv2026

Towards Fine-Grained Multi-Dimensional Speech Understanding: Data Pipeline, Benchmark, and Model

While speech Large Language Models (LLMs) excel at conventional tasks like basic speech recognition, they lack fine-grained, multi-dimensional perception. This deficiency is evident in their struggle to disentangle complex features like micro-acoustic cues, acoustic scenes, and paralinguistic signals. This resulting incomplete comprehension of real-world speech fundamentally bottlenecks the development of perceptive and empathetic next-generation speech systems. At its core, this persistent perceptual limitation primarily stems from three interacting factors: scarce high-quality expressive data, absent fine-grained modeling for multi-dimensional attributes, and reliance on restricted coverage, coarse-grained benchmarks. We address these challenges through three pillars: First, our robust data curation pipeline resolves complex acoustic environments and long-audio timestamp alignment challenges to extract a high-quality spontaneous speech corpus from audiovisual sources. Second, we construct FMSU-Bench, a pioneering benchmark covering 14 speech attribute dimensions to rigorously assess the fine-grained, multi-dimensional speech understanding capabilities of current models. Third, empowered by our curated corpus, we introduce FM-Speech. Driven by a decoupled attribute modeling and progressive curriculum fine-tuning framework, it substantially elevates fine-grained, multi-dimensional acoustic perception. Extensive evaluations on FMSU-Bench reveal that current speech LLMs still require significant improvement in multi-dimensional, fine-grained understanding. In contrast, FM-Speech substantially outperforms current open-source models, establishing a robust paradigm for real-world speech understanding.

eess.AS↗
arXiv2026

Tensor network investigation of the monomer-dimer model on the honeycomb lattice

The monomer-dimer model is one of the most well-known unsolved lattice models. In this paper we study the monomer-dimer model on the honeycomb lattice using the tensor network method, in the case that the dimer and monomer activities are 1. The monomer-dimer configurations are exactly mapped into the ground states of the antiferromagnetic Ising model on the Kagomé lattice in the critical field $H_{\rm{ex}}=4J$, and the tensor network is constructed based on the local ground states of each Ising triangle. The VUMPS approach is employed to contract the tensor network, providing a high-precision result of the monomer-dimer problem. We also revisit the edge coloring problem on the honeycomb lattice and discuss its relationship to the monomer-dimer model. Finally we formulate the monomer-dimer problem in the language of the sixteen-vertex model, and discuss the non-integrability of the general monomer-dimer model and the integrability of the pure dimer model.

cond-mat.stat-mech↗