Searcharxiv⌕ Search

arXiv subjects

Zhen Chen

Publications and source records attributed to Zhen Chen.

At least 19 recordsLinked to original sources
arXiv2026

GroundSeek-CXR: Retrospective Evaluation-Guided Active Observation Framework for Multi-Label Chest Radiography Diagnosis

Chest radiography (CXR) diagnosis is critical for thoracic disease assessment, yet challenged by the need to identify subtle lesions and co-occurring abnormalities across regions. While recent vision-language models can actively observe local regions, existing preference-based methods supervise each observation by its immediate effect on the current diagnosis, failing to capture how each observation and stopping decision contributes to the final diagnosis. To address this issue, we propose GroundSeek-CXR, an active CXR diagnosis framework that learns observation and stopping decisions by retrospectively evaluating them with the final diagnosis. Specifically, we first devise a Retrospective Diagnostic Evaluation (RDE) that scores observation and stop candidates from offline Monte Carlo tree search through independent continuations. Moreover, we design a Retrospective Gap-weighted Preference Optimization (RGPO) that learns from retrospective-score preferences weighted by their score gaps, prioritizing actions that improve diagnostic accuracy. Extensive experiments demonstrate that our GroundSeek-CXR framework outperforms state-of-the-art methods, and improves abnormal-image sample-F1 by 4.48% over immediate-score preference learning.

cs.CV↗
arXiv2026

SurgEye: Instruction Tuning-Free Perception-Reasoning Framework for Surgical Scene Understanding

Surgical scene understanding is pivotal for computer-assisted intervention, yet demands compositional reasoning across instruments, anatomy, and workflow. Existing surgical vision-language models rely on instruction tuning with large-scale question-answer pairs, entailing laborious data curation and costly retraining. To bridge this gap, we propose SurgEye, a perception-reasoning framework that grounds reasoning in a vision foundation model without instruction tuning. Specifically, we devise Routed Multi-Task Perception (RMP) that routes the vision foundation model through task-specific pathways learned from public perception labels, producing spatial and temporal evidence. We further introduce Agentic Compositional Reasoning (ACR) that employs a frozen large language model to coordinate RMP specialists and temporal memory as callable tools, composing evidence across tasks and time into grounded answers. Moreover, we construct SurgEye-Bench spanning eight surgical procedures, pairing each question with perception labels for verifiable evaluation. Extensive experiments demonstrate that SurgEye outperforms state-of-the-arts in surgical reasoning. The source code will be available.

cs.CV↗
arXiv2026

SurgStream: A Workflow-Anchored Proactive Streaming Framework and Benchmark for Real-Time Surgical Video Understanding

Surgical video understanding is essential for intraoperative assistance such as workflow monitoring, instrument preparation and procedural documentation. Delivering such assistance in real time requires a model to jointly perceive the ongoing scene, remember past surgical events, and respond before clinically relevant moments pass. Existing approaches provide limited support for jointly retaining procedural evidence and deciding when to respond in streaming surgical video. In this work, we propose SurgStream, a workflow-anchored proactive streaming framework for real-time surgical video understanding. We first introduce Workflow-Anchored Memory (WAM) to record structured instrument-action-target observations and fold them into phase-level summaries, providing compact and queryable surgical memory. Deadline-aware Transition Learning (DTL) is then devised to supervise response states within clinical windows, enabling timely responses before procedural deadlines. We further introduce SurgStream-Bench, comprising 11.7K streaming interactions across various tasks covering real-time perception, backward memory queries, forward proactive QA, proactive alerts and caption memory. Experiments on SurgStream-Bench show gains over state-of-the-art methods, especially in memory and forward proactive QA. The source code will be available.

cs.CV↗
arXiv2026

Radiology Guideline-derived Decision Graph for Efficient Clinical Decision Support

Radiological risk stratification and management planning are essential downstream tasks in medical imaging. Automating these tasks is challenging because clinical guideline criteria are distributed across prose, tables, and cross-referenced definitions. We present Guideline-derived Decision Graph (GDGraph), a framework that decouples guideline compilation from case evaluation. During compilation, a large language model partitions guidelines into section-based chunks, extracts local decision subgraphs with linked definitions, and aggregates related branches into an executable decision graph. During inference, a fast System One model answers localized clinical questions while deterministic traversal resolves Boolean logic, turning opaque single-pass predictions into a traceable decision path without task-specific training. Across 347 cases spanning renal cyst, adnexal mass, and thyroid nodule guidelines, GDGraph improves System One accuracy from 57.3% to 90.8%, outperforming retrieval-augmented GPT-4o (80.1%) and graph-guided GPT-4o with chain-of-thought (82.7%) at substantially lower latency and cost. The source code will be available.

cs.CV↗
arXiv2026

HiMA-Ecom: A Long-Horizon Memory Benchmark for Hierarchical Multi-Agent E-commerce Assistants

Long-horizon agentic systems become long-context systems by accumulation rather than through a single long document, since every step reassembles a prompt from role instructions, tool and sub-agent descriptions, recalled memories, dialogue history, and previous tool outputs, separately for every agent. This paper studies how such accumulated context can be trained, evaluated, and managed in hierarchical multi-agent assistants. We present HiMA-Ecom, a hierarchical multi-agent benchmark for e-commerce assistants with 22.8K instances, of which 17.7K carry memory entries in context, pairing agent-specific supervised fine-tuning data with system-level supervision for end-to-end multi-agent reinforcement learning. Building upon it, HiMA-R1 keeps the context budget of a trajectory under control through Variance-Reduction Group Relative Policy Optimization (VR-GRPO), which samples along an initial trajectory so that sampling cost grows additively with trajectory length and updates the agents with the largest reward variance, together with an efficiency reward on trajectory length and a memory evolution mechanism that repurposes GRPO rewards as cost-free signals for what stays in context. Agents built upon 3B and 7B open-source models match DeepSeek-R1 and surpass DeepSeek-V3 by 6.0 points on average, while reward-driven memory reduces the update steps needed for convergence by roughly 55%.

cs.AI↗
arXiv2026

Role-Augmented Intent-Driven Generative Search Engine Optimization

Generative search engines (GSEs), which combine large language models (LLMs) with retrieval-augmented generation (RAG), are reshaping how users access Web information and making source visibility and influence in generated responses new optimization objectives. Existing generative search engine optimization (G-SEO) methods seek to improve content visibility, but many rely on direct access to target queries or GSE feedback elicited by those queries. However, Web content is typically optimized and published before future user queries are known, making G-SEO without access to target queries or their associated feedback an important yet underexplored problem. To address this, we propose RAID, a Role-Augmented Intent-Driven G-SEO method that operates without access to target queries or query-dependent GSE feedback. RAID infers a primary retrieval intent from source content, refines it through 4W multi-role deep reflection on prospective searchers' information needs, and uses both the primary and refined intents to jointly plan and rewrite source content. We further extend GEO-BENCH with diverse topic-related query variants, using both the original queries and their variants exclusively for evaluation to assess generalization to queries unseen during optimization. Across 100 samples and 500 evaluation queries, RAID achieves Relative Improvement Scores (RIS) of 3.00\% for Objective Overall and 1.32\% for Subjective Average, exceeding the best corresponding results among the nine predefined GEO rewriting strategies by 1.31 and 0.74 percentage points, respectively. With an optimization considered effective when $\mathrm{RIS}>0$, its effective optimization rates for Objective Overall and Subjective Average average 58.0\%, exceeding the highest corresponding average among the predefined strategies by 4.5 percentage points.

cs.IR↗
arXiv2026

LRMoE: Demand-Side Load Re-Calibration for Efficient MoE Training on Memory-Constrained Nodes

Sparse Mixture-of-Experts (MoE) models offer scalable capacity for Web applications, but uneven routing creates computation and communication bottlenecks. Existing placement and replication methods redistribute this workload, yet additional expert copies compete with token-related buffers for limited GPU memory. We present LRMoE, a memory-aware MoE training system that jointly recalibrates backward demand and balances the retained workload. During forward execution, it plans retention quotas and expert placement under resource constraints while preserving all routed computations. During backpropagation, a load recalibration metric estimates each token-expert pair's contribution to the expert update and selects pairs within these quotas. A capacity-aware routing algorithm distributes retained pairs among eligible instances of their original experts. Physical compaction then forms smaller communication payloads and variable-size grouped GEMMs, reducing backward traffic, computation, and execution-buffer demand. Experiments on Qwen3-30B-A3B and GLM4.5-106B-A12B show that LRMoE reduces cumulative MoE-layer training time to fixed-loss targets by 21.2-44.8% relative to SYMI.

cs.AI↗
arXiv2026

Amplitude-Only FFN Intervention for Tool-Structured LLM Inference: Method, Gated Evaluation Protocol, and Cross-Model Empirical Results

We study amplitude-only rescaling of feed-forward network (FFN) activations to repair tool calls and other structured outputs, using multiplicative coordinate masks rather than additive steering vectors. Amplitude Gating (AG) keeps pretrained weights fixed and applies magnitude/entropy-based masks at six SwiGLU sites, with a Newton-Schulz windowed ranking variant. Across three Qwen checkpoints and six text/tool endpoints, effects are task dependent. Under frozen local scorers, Qwen3.5-9B non-NS configurations improve two function-call endpoints by +2.61/+2.67 percentage points (pp) and static ToolBench in/out-domain by +1.64/+2.82 pp; JSON gains are small. These gains survive comparing the weaker of two observed AG runs with the stronger baseline, supporting observed-run robustness rather than population significance. Scores measure lexical or answer/action-name agreement, not executed-tool success. Under the current experimental conditions, the six-task random-forest refit improves native score by +0.46 pp with a positive conditional paired interval; transfer to other settings requires further evaluation. The results support task-specific interventions; causal mechanisms, equal-budget advantages, repeated-run uncertainty, and serving cost require further evaluation. Code and experimental data: https://doi.org/10.5281/zenodo.23654706.

cs.CL↗
arXiv2026

FlexMARL: Graph-Driven Rollout-Training Co-Design for LLM-Based Multi-Agent Reinforcement Learning

Despite algorithm-level innovations for multi-agent reinforcement learning (MARL), the underlying infrastructure for large-scale MARL remains underexplored. Existing training frameworks primarily optimize for single-agent scenarios and fail to capture the graph-structured nature of MARL. This fundamental mismatch leads to three MARL-specific challenges: topology-induced rollout skew, critical path synchronization barriers, and dynamic training demand. To bridge this gap, we propose FlexMARL, a graph-driven rollout-training co-designed framework for large-scale LLM-based MARL. FlexMARL introduces the multi-agent execution graph (MEG) and materializes the intermediate representation as the MEG Store, which provides decentralized multi-table storage and serves as the data hub for system components. Built around the MEG Store, the Rollout Engine develops dependency parallel sampling with hierarchical load balancing, which reduces inference latency and balances intra-agent/inter-agent loads. The Joint Orchestrator exploits an agent-level asynchronous pipeline to mitigate the critical path synchronization barriers while maintaining consistency. The Training Engine achieves on-demand resource allocation by collectively managing training processes for individual agents. The training states of different agents are swapped via location-agnostic communication APIs. Empirical results on a production cluster demonstrate that FlexMARL achieves up to 7.3x speedup and improves hardware utilization by up to 5.6x compared to existing frameworks.

cs.LG↗
arXiv2026

Counterexamples to Two Conjectures on Tangles in Matroid Minors

We give counterexamples to two conjectures of Brettell, Jowett, Oxley, Semple, and Whittle concerning tangles in matroid minors. We construct a matroid $P_0$ with a tangle $\mathcal T$ of order seven. For each $i\in\{4,5,6,7\}$, let $\mathcal T_i$ denote its order-$i$ truncation. No proper minor $N$ of $P_0$ admits, for every such $i$, a tangle of order $i$ whose breadth equals that of $\mathcal T_i$, and $P_0$ is not weakly $4$-connected. We also construct a matroid $P_1$ with a tangle of order four and breadth six, but no internally $4$-connected minor of $P_1$ admits a tangle of order four.

math.CO↗
arXiv2026

Rethinking Calibration for High-Sparsity LLM Pruning: From Single Sources to Capability Coverage

Post-training LLM pruning relies on a small calibration set, usually chosen heuristically and judged by WikiText perplexity or by mean accuracy on general-domain commonsense benchmarks. This single-axis view hides capability-specific effects: at 60\% sparsity, strategies separated by $2.85$ points in averaged commonsense accuracy diverge by $51.9$ points in code retention, and across 15 sources no single one preserves all capabilities. We treat calibration as a capability-coverage problem. Balanced, a uniform four-way mixture under a fixed budget, tests the principle on real corpora: it outperforms every evaluated single source, and its margin over C4 grows from $+4.6$ points at 30\% sparsity to $+18.8$ at 60\%. The advantage holds across model scale and a second model family, and transfers to held-out benchmarks disjoint from calibration construction. When aligned corpora are unavailable, IGSP approximates that mixture from the base model and the evaluation taxonomy alone, outperforming prior self-generation methods by $7.23$--$12.79$ points on the LLaMA main suite. On Qwen2.5-7B, IGSP exceeds Balanced on both the main suite and held-out benchmarks. Capability coverage is a first-order design variable for high-sparsity pruning.

cs.LG↗
arXiv2026

GRASE: Geometry-native Real-to-sim Agentic Scene Modeling from Video

Real-to-sim conversion aims to turn a captured video of a real environment into an editable 3D scene while preserving the geometry and layout of what was observed. Existing methods often follow a one-pass, loosely coupled paradigm: each stage receives only limited feedback from previous stages, allowing errors to propagate without reliable correction and leaving the final scene inconsistent. We present GRASE, an agentic framework that closes this loop by treating the captured multi-view video observations as a standing reference throughout scene modeling. Instead of committing to intermediate estimates, GRASE revises its current hypothesis whenever rendered evidence disagrees with the capture, overcoming the fixed intermediate assumptions of conventional pipelines. Specifically, GRASE first fits the architectural shell and its openings from depth and multi-view cues. An agent then places generated assets in a shared coordinate frame while checking their spatial relationships. Finally, each asset's geometry and placement are refined iteratively to improve agreement across the observed views. By coupling these steps through feedback, GRASE localizes structural discrepancies and revises only the responsible components, while preserving cross-view consistency. The resulting 3D scenes are faithful to the input, consistent across views, and suitable for simulations. In our experiments, GRASE substantially outperforms other state-of-the-art methods, including a direct real-to-sim conversion using GPT6-Astra.

cs.CV↗
arXiv2026

The optimality of an (s, S) hiring policy on a workforce planning problem with fixed recruitment costs and binomial turnover

We study a finite-horizon nonstationary workforce planning problem in which staff turnover in each period follows a binomial distribution whose parameters depend on the post-hiring workforce level. The model incorporates a fixed hiring cost that is incurred whenever recruitment occurs, regardless of the number of employees hired; this represents explicit (e.g. administration costs) or implicit (e.g. team disruption) factors that are commonplace in workforce planning. The objective is to minimise the expected total cost, including recruitment, salary, and shortage costs, where deviations below period-specific staffing requirements are penalised. To analyse this stochastic dynamic program with decision-dependent transition probabilities, we establish the discrete convexity of the variable single-period cost (the sum of expected salary and penalty costs) and the K-convexity of the expected total cost in presence of endogenous uncertainty. We then show that the optimal hiring policy exhibits an (s, S)-type structure: when the workforce level in a given period falls below a threshold s, staff are hired up to level S; otherwise, no hiring occurs. To compute approximate policy parameters, we develop a mixed-integer linear programming heuristic based on a piecewise approximation. In our computational study, we demonstrate that our approach achieves fast computation and small optimality gaps on the test instances.

math.OC↗
arXiv2026

Unravel: Recovering an Interactive 3D Scene from a Single Image

We present Unravel, a training-free framework that turns a single image into an interactive 3D scene, where every object is an independent and movable asset. Existing generators either produce one entangled geometry in which no object can be moved, or predict all objects in one feed-forward pass, where a wrong guess cannot be detected or repaired. Unravel instead takes a different approach: it first takes the scene apart, and then puts it back together. Starting from the input image, an agentic harness that iteratively peels the image layer by layer, and adopts point cloud as major modality for spatial relationship recovery. Specifically, a vision-language model plans which objects are fully visible, an image editor removes them and restores the occluded regions. Every edit is verified before being incorporated into the reconstruction. This process gradually turns occluded parts of the scene into direct observation, allowing objects to be reconstructed when they are best exposed rather than hallucinated from the original image alone. The scene is then rebuilt in reverse removal order, so that the supporting objects are naturally placed before the objects they support. Since a single-view image only reveals part of an object, we treat the observed points as a lower bound on its size: an asset may grow to cover its points, but can not shrink away from them. Finally, we physically settle the reconstructed scene to remove floating objects and interpenetrations, followed by an agentic audit that corrects remaining placement errors. Extensive experiments on diverse indoor scenes show that Unravel recovers more objects, places them more faithfully, and produces scenes that are more physically-plausible than existing methods.

cs.CV↗
arXiv2026

CompileMask: Bridging the Gap Between Expressiveness and Efficiency in Attention Masking

Modern attention workloads require diverse masking patterns whose boundaries depend on runtime metadata, yet efficient attention kernels natively support only a small set of patterns. Existing methods for supporting custom masks in these kernels face an expressiveness-efficiency trade-off. Predicate-based methods are expressive but incur substantial preprocessing and execution overhead, whereas interval-based methods execute efficiently but restrict expressiveness through a predefined interval count. We present CompileMask, which addresses this trade-off by connecting expressive mask specifications with efficient attention execution. We design a compiler that analyzes visibility rules during the offline stage and generates a mask-specific preprocess kernel. We further propose MaskSpec, a region-based DSL that lets users express these rules. During the online stage, the generated kernel reads batch metadata and produces a compact Execution Descriptor that lets the attention kernel efficiently check visibility by comparing element indices against precomputed endpoints. Experiments show that CompileMask achieves up to $3.8\times$ speedup over state-of-the-art methods.

cs.DC↗
arXiv2026

FissionGuard: Testing Local Watermark Proportions with FDR Control

A watermark detector can flag a text even when only a small fraction of its tokens are watermarked. We study sequential auditing to identify local passages whose watermark proportions exceed a prespecified threshold. Each alarm fixes a candidate passage's endpoint, while its start is chosen from earlier tokens. Reusing the same evidence to locate and test a passage can invalidate inference. We propose FissionGuard, which uses data fission to construct an alarm value and a validation value from each token-level $p$-value. The alarm values guide alarms and localization. The validation values yield passage-level $p$-values through partial conjunction, excluding as many of the strongest signals as the proportion null allows. These $p$-values are used to test the corresponding hypotheses at levels assigned by an online multiple-testing procedure. Under the stated watermark generation model, allowing dependence among token-level $p$-values at watermarked positions, we prove finite-sample false discovery rate (FDR) control among reported passages at every fixed finite horizon. We also derive a finite-sample lower bound on power for testing nonnull candidates. Experiments on synthetic streams and Mistral and Qwen generations across four marking layouts show gains in watermarked-token coverage over sample splitting, with empirical FDR below the target.

stat.ME↗
arXiv2026

Do Self-Evolving Skills Generalize to Held-Out Tasks?

AI agents can externalize what they learn from past tasks into reusable \emph{skills}, such as procedures, checklists, code, or other executable artifacts, that can be retrieved and reused when solving new tasks. Self-evolving skill methods keep rewriting these skills after each round of practice on training tasks, and the skill is then used on new tasks of the same kind. We ask a question: does the improvement a skill shows on its training tasks carry over to new test tasks? We test five self-evolving methods and a one-shot skill on six benchmarks, with the same model, the same agent, and the same train/test split for every method. Of the 21 skills that improve on their training tasks, 5 keep all of that improvement on the test tasks, 13 keep part of it, and 3 keep none of it. No existing method is best everywhere. When we read the skills, the ones that carry over badly often fix details that should depend on the task, such as column names and output files, or turn a fix for one failure into a rule for every task. An LLM judge that reads the skill content can often see this: it ranks finished skills the same way the test results do in 86\% of pairs. But it predicts the effect of a single edit poorly, so edits still have to be tested by running them. Based on these findings, we describe Generalizable Skill Optimization (GSO), which keeps only a guide for writing skills and writes a new skill for each task; it scores highest on all six benchmarks.

cs.AI↗
arXiv2026

Regulating oxygen content and superconductivity in La$_3$Ni$_2$O$_{7+δ}$

The synthesis of high-quality Ruddlesden-Popper (RP) nickelates remains challenging due to variations in oxygen content and the prevalence of intergrown RP phases. Precisely controlling the stoichiometry and characterizing the resulting physical properties are essential for understanding the mechanism of high-$T_c$ superconductivity in these materials. In this work, we synthesize a series of La$_3$Ni$_2$O$_{7+δ}$ samples with systematically controlled oxygen content and perform comprehensive structural and compositional analyses. Precise oxygen tuning enables us to tailor the microstructure, yielding a pure bilayer phase, a mixture of bilayer and hybrid single-layer-bilayer phases, and a predominantly bilayer phase containing trilayer intergrowths. High-pressure transport measurements reveal distinct superconducting transitions with contrasting $T_c$ values, corresponding to the bilayer phase, the hybrid phase, and trilayer inclusions. Notably, we find that oxygen content not only governs the phase purity$-$i.e., the presence of intergrowth phases$-$but also directly modulates the upper critical field ($H_{c2}$) of the bilayer superconductivity. By establishing a phase diagram of $T_c$ and $H_{c2}$ as functions of oxygen content in La$_3$Ni$_2$O$_{7+δ}$, this work advances synthetic control and provides new insights into the superconducting mechanism of RP nickelates.

cond-mat.supr-con↗