Search arXiv⌕ Search

arXiv subjects

Chen Luo

Publications and source records attributed to Chen Luo.

At least 19 recordsLinked to original sources
arXiv2026

Subspace Alignment for Vision-Language Model Test-time Adaptation

Vision-language models (VLMs), despite their extraordinary zero-shot capabilities, are vulnerable to distribution shifts. Test-time adaptation (TTA) emerges as a predominant strategy to adapt VLMs to unlabeled test data on the fly. However, existing TTA methods heavily rely on zero-shot predictions as pseudo-labels for self-training, which can be unreliable under distribution shifts and misguide adaptation due to two fundamental limitations.First (Modality Gap), distribution shifts induce gaps between visual and textual modalities, making cross-modal relations inaccurate. Second (Visual Nuisance), visual embeddings encode rich but task-irrelevant noise that often overwhelms task-specific semantics under distribution shifts. To address these limitations, we propose SubTTA, which aligns the semantic subspaces of both modalities to enhance zero-shot predictions to better guide the TTA process. To bridge the modality gap, SubTTA extracts the principal subspaces of both modalities and aligns the visual manifold to the textual semantic anchor by minimizing their chordal distance. To eliminate visual nuisance, SubTTA projects the aligned visual features onto the task-specific textual subspace, which filters out task-irrelevant noise by constraining visual embeddings within the valid semantic span, and standard TTA is further performed on the purified space to refine the decision boundaries. Extensive experiments on various benchmarks and VLM architectures demonstrate the effectiveness of SubTTA, yielding an average improvement of 2.24% over state-of-the-art TTA methods. Our code is available at https://github.com/zhichenz98/SubTTA_EMNLP26.

cs.CV↗
arXiv2026

PACEShop: Evaluating Personalized, Actionable, Compositional, and Evidence-grounded Shopping Assistants

Shopping assistants are shifting from ranked product lists toward structured decision support, where systems must synthesize shopper context, product evidence, and next-step guidance into a coherent recommendation experience. This changes the unit of evaluation: a fluent response can still fail by ignoring shopper context, contradicting itself across components, or leaving defects too vague to localize. Existing personalization, grounding, and LLM-as-a-judge benchmarks cover pieces of this problem, but they do not define a joint evaluation target for structured shopping-assistant responses. We formulate this missing evaluation target as PACE: Personalized, Actionable, Compositional, and Evidence-grounded evaluation. We instantiate PACE with two artifacts: PACEShop, a benchmark dataset that makes the target measurable through 22,625 controlled records with structured personas, auditable evidence pools, GOOD/BAD labels, and gold defect family and location annotations; and PACEJudge, a training-free judging protocol that makes the target reportable through a structured output contract. Our experiments show that generic judges can recognize broad quality but fail to recover the diagnostic fields required for PACE; PACEShop makes these failures verifiable, and PACEJudge improves persona-source, cross-component, grounding, and family/location closure without retraining, showing that realistic shopping-assistant evaluation requires a task-matched output contract rather than only a stronger backbone or scalar prompt.

cs.CL↗
arXiv2026

Vector-field control and emergent basal-plane anisotropy of magnetic textures in noncentrosymmetric (Fe$_{0.63}$Ni$_{0.3}$Pd$_{0.07}$)$_3$P

(Fe$_{0.63}$Ni$_{0.3}$Pd$_{0.07}$)$_3$P is a room-temperature magnet with $S_4$ symmetry that hosts a rich variety of topological spin textures. Here, we report a combined resonant small-angle x-ray scattering and ptychography study of (Fe$_{0.63}$Ni$_{0.3}$Pd$_{0.07}$)$_3$P in vector magnetic fields over a broad temperature range. We demonstrate deterministic vector-field control of magnetic stripe domains, where in-plane fields continuously rotate their orientation via a transition from a chiral stripe to an achiral fan configuration. Furthermore, at 50 K and below, the stripe orientation becomes metastably pinned and retains its field-trained direction. While the magnitude of the wavevector is nearly isotropic within the basal plane at room temperature, a pronounced temperature evolution of anisotropic interactions emerges upon cooling. In particular, non-trivial anisotropy axes develop at 20-50 K reflecting the combined effects of magnetocrystalline anisotropy, anisotropic exchange, and Dzyaloshinskii-Moriya interaction (DMI), whose effective orientation is found to rotate with temperature. These results establish (Fe$_{0.63}$Ni$_{0.3}$Pd$_{0.07}$)$_3$P as a model system for vector-field control of chiral spin textures and reveal a previously unrecognized temperature-driven evolution of the effective DMI landscape in a noncentrosymmetric magnet.

cond-mat.mes-hall↗
arXiv2026

Proximity-Induced Skyrmion Stabilization at the Cu2OSeO3/Bi2Se3 Interface

We investigate proximity-induced magnetic interactions at the interface between the topological insulator Bi2Se3 and the chiral magnetic insulator Cu2OSeO3, with particular focus on the low temperature skyrmion phase. Broadband ferromagnetic resonance spectroscopy reveals enhanced stability of noncollinear spin textures in the Cu2OSeO3/Bi2Se3 heterostructure compared with bare Cu2OSeO3. In addition to an extra resonance mode in the tilted conical phase that is absent in bare Cu2OSeO3, field cycling resolves two counterclockwise skyrmion resonance branches separated by approximately 238 MHz, consistent with the coexistence of a bulk skyrmion lattice and an interfacial skyrmion phase stabilized by proximity-induced exchange coupling and enhanced interfacial Dzyaloshinskii-Moriya interactions. The finite frequency separation indicates that the two skyrmion phases occupy distinct magnetic energy landscapes while retaining similar resonance character. Resonant elastic x-ray scattering measurements further confirm that the interfacial skyrmion phase spans a broader magnetic-field range than the bulk phase, demonstrating enhanced stability and ordering of topological spin textures at the interface. These findings establish interface engineering as a promising route for extending the stability regime of skyrmion and tilted-conical phases in topological-magnetic heterostructures.

cond-mat.mtrl-sci↗
arXiv2026

SENTINEL: Failure-Driven Reinforcement Learning for Training Tool-Using Language Model Agents

Language model agents are increasingly effective in solving realistic tasks through multi-turn tool use. However, training reliable tool-using agents remains challenging in practice. While reinforcement learning provides an on-policy paradigm for improving agents from their own environment interactions, its effectiveness depends heavily on the training task distribution. When tasks are fixed before training, the task distribution can become increasingly mismatched with the policy's evolving capabilities, causing many rollouts to be spent on uninformative tasks. We propose SENTINEL, a failure-driven reinforcement learning framework that turns the Solver's rollout failures into targeted training tasks. SENTINEL follows a Controller--Proposer--Solver loop: the Controller analyzes failed trajectories and summarizes recurring error patterns, the Proposer generates executable tasks that stress these weaknesses, and the Solver is trained on the targeted tasks. On Tau2-Bench Retail with Qwen3-4B-Thinking-2507, SENTINEL improves Pass\^{}1 from 66.4 to 74.9 and outperforms RL on general synthetic tasks across Pass\^{}k metrics. These results demonstrate that model failures provide an effective and scalable source of targeted training signal for improving tool-using language model agents.

cs.CL↗
arXiv2026

The World Won't Stay Still: Programmable Evolution for Agent Benchmarks

LLM-powered tool-calling agents fulfill user requests by interacting with environments, querying data, and invoking tools in a multi-turn process. Yet, most existing benchmarks evaluate these systems under static environment interfaces, with fixed schemas and toolsets, making it difficult to assess how agents behave as environments evolves -- when capabilities are added, reorganized, or deprecated across successive environment versions. In this paper, we study structured environment evolution as a benchmark-construction problem for tool-calling agents. We propose ProEvolve, a graph-based framework that makes environment evolution programmable. At its core, a typed relational graph provides a unified, explicit representation of the environment - data, tools, and schema. Under this formalism, adding, removing, or modifying capabilities are expressed as graph transformations that coherently propagate updates across tools, schemas, and data access. Building on this, ProEvolve supports (1) automatic generation of evolved executable environments through explicit graph transformations, and (2) graph-grounded construction of task sandboxes via subgraph sampling and instantiation. We validate ProEvolve in two tool-calling domains, e-commerce and airline booking, in terms of quality, implementation validity, and failure modes. Finally, we use the generated benchmark as a downstream diagnostic to study how representative agents behave under structured environment evolution.

cs.AI↗
arXiv2026

WEBSERV: A Full-Stack and RL-Ready Web Environment for Training Web Agents at Scale

Reinforcement learning (RL) for web agents demands environments that are both effective for evaluation and efficient enough for large-scale on-policy training. Current web environments fall short: server-side Docker setups are too resource-intensive for massive parallel rollouts, while browser-side interfaces produce noisy observations, execute actions unreliably under modern single-page applications, and omit visual interactivity cues. We introduce WebServ, a full-stack, RL-ready web environment that addresses these limitations end-to-end. On the server side, WebServ uses Incus containers with block-level copy-on-write, reducing launch latency by ~5x and persistent storage by ~240x, enabling 200+ concurrent isolated environments on a single host. On the browser side, WebServ provides a compact, site-agnostic observation and action interface derived automatically from the DOM with human-aligned interactivity cues, and a robust action execution backend using network-aware waiting for reliable SPA support. On WebArena-Lite, WebServ achieves state-of-the-art single-prompt results, with controlled comparisons confirming consistent gains across GPT-4o, OpenAI-o3, and Llama-3.1-8B over vanilla WebArena. We further train Qwen3-4B and Qwen3-30B-A3B with RL entirely within WebServ; the RL-trained 4B model achieves 55.5% mean accuracy, surpassing both Claude 4.5 Sonnet (50.0%) and the RL-trained 8B model from WebAgent-R1 (51.8%).

cs.LG↗
arXiv2026

Firefly: Illuminating Large-Scale Verified Tool-Call Data Generation from Real APIs

Training tool-calling agents requires large-scale trajectory data with verifiable labels, yet existing approaches either synthesize environments that diverge from real API behavior or generate tasks without ground-truth outcomes for verification. We present FireFly, a pipeline for generating verified tool-call data from real-world MCP servers. Our key insight is to invert the standard synthesis pipeline: rather than generating tasks and hoping they are solvable, we first let a strong LLM explore real APIs along graph-guided DAG structures, then synthesize tasks backward from observed outcomes, guaranteeing label correctness by construction. To handle the scale of real-world tool spaces (${\sim}$1,000 tools), we build a pairwise tool graph and sample sub-DAGs to focus exploration on semantically coherent workflows. To address environment drift in live APIs, we construct a retrieval-augmented simulator that caches all exploration results and replays them during training and evaluation, enabling fully offline and reproducible RL. Applying this pipeline yields 5,144 verified tasks spanning 240 servers and 993 tools. A 4B-parameter model trained with GRPO on FireFly matches Claude Sonnet 4.6 on our held-out test set and shows improvements on multiple tool-calling benchmarks including Tau2-Bench, MCPMark, and MCP-Atlas.

cs.SE↗
arXiv2026

A Surface-confined Spiral State With the Double Period in the Cubic Chiral Helimagnet Cu$_2$OSeO$_3$

The chiral magnetoelectric insulator Cu$_2$OSeO$_3$ hosts a rich and anisotropic magnetic phase diagram that includes helical, conical, field-polarised, tilted conical, and skyrmion lattice phases. Using resonant elastic x-ray scattering (REXS), we uncover a new spiral state confined to the surface of Cu$_2$OSeO$_3$. This surface-confined spiral state (SSS) displays a real-space pitch of $\sim$120 nm, which remarkably is twice the length of the incommensurate structures observed to-date in Cu$_2$OSeO$_3$. The SSS phase emerges at temperatures below 30~K when the magnetic field is applied between $3^\circ$ to $18^\circ$ away from the $\langle\text{110}\rangle$ crystallographic axes. Its surface localisation is demonstrated through a combination of REXS in reflection and transmission geometries, with complementary small-angle neutron scattering measurements suggesting its absence from the bulk. We attribute the stabilisation of the SSS to competing anisotropic interactions at the crystal surface. The discovery of a robust, surface-confined spiral paves the way for engineering energy-efficient, nanoscale spin-texture platforms for next-generation devices.

cond-mat.str-el↗
arXiv2026

Trajectory2Task: Training Robust Tool-Calling Agents with Synthesized Yet Verifiable Data for Complex User Intents

Tool-calling agents are increasingly deployed in real-world customer-facing workflows. Yet most studies on tool-calling agents focus on idealized settings with general, fixed, and well-specified tasks. In real-world applications, user requests are often (1) ambiguous, (2) changing over time, or (3) infeasible due to policy constraints, and training and evaluation data that cover these diverse, complex interaction patterns remain under-represented. To bridge the gap, we present Trajectory2Task, a verifiable data generation pipeline for studying tool use at scale under three realistic user scenarios: ambiguous intent, changing intent, and infeasible intents. The pipeline first conducts multi-turn exploration to produce valid tool-call trajectories. It then converts these trajectories into user-facing tasks with controlled intent adaptations. This process yields verifiable task that support closed-loop evaluation and training. We benchmark seven state-of-the-art LLMs on the generated complex user scenario tasks and observe frequent failures. Finally, using successful trajectories obtained from task rollouts, we fine-tune lightweight LLMs and find consistent improvements across all three conditions, along with better generalization to unseen tool-use domains, indicating stronger tool-calling ability.

cs.CL↗
arXiv2026

SessionIntentBench: A Multi-task Inter-session Intention-shift Modeling Benchmark for E-commerce Customer Behavior Understanding

Session history is a common way of recording user interacting behaviors throughout a browsing activity with multiple products. For example, if an user clicks a product webpage and then leaves, it might because there are certain features that don't satisfy the user, which serve as an important indicator of on-the-spot user preferences. However, all prior works fail to capture and model customer intention effectively because insufficient information exploitation and only apparent information like descriptions and titles are used. There is also a lack of data and corresponding benchmark for explicitly modeling intention in E-commerce product purchase sessions. To address these issues, we introduce the concept of an intention tree and propose a dataset curation pipeline. Together, we construct a sibling multimodal benchmark, SessionIntentBench, that evaluates L(V)LMs' capability on understanding inter-session intention shift with four subtasks. With 1,952,177 intention entries, 1,132,145 session intention trajectories, and 13,003,664 available tasks mined using 10,905 sessions, we provide a scalable way to exploit the existing session data for customer intention understanding. We conduct human annotations to collect ground-truth label for a subset of collected data to form an evaluation gold set. Extensive experiments on the annotated data further confirm that current L(V)LMs fail to capture and utilize the intention across the complex session setting. Further analysis show injecting intention enhances LLMs' performances.

cs.CL↗
arXiv2026

CARAMEL: A Succinct Read-Only Lookup Table via Compressed Static Functions

Lookup tables are a fundamental structure in many data processing and systems applications. Examples include tokenized text in NLP, quantized embedding collections in recommendation systems, integer sketches for streaming data, and hash-based string representations in genomics. With the increasing size of web-scale data, such applications often require compression techniques that support fast random $O(1)$ lookup of individual parameters directly on the compressed data (i.e. without blockwise decompression in RAM). While the community has proposd a number of succinct data structures that support queries over compressed representations, these approaches do not fully leverage the low-entropy structure prevalent in real-world workloads to reduce space. Inspired by recent advances in static function construction techniques, we propose a space-efficient representation of immutable key-value data, called CARAMEL, specifically designed for the case where the values are multi-sets. By carefully combining multiple compressed static functions, CARAMEL occupies space proportional to the data entropy with low memory overheads and minimal lookup costs. We demonstrate 1.25-16x compression on practical lookup tasks drawn from real-world systems, improving upon established techniques, including a production-grade read-only database widely used for development within Amazon.com.

cs.DS↗
arXiv2026

Prune as You Generate: Online Rollout Pruning for Faster and Better RLVR

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Large Language Models (LLMs). However, methods such as GRPO and DAPO suffer from substantial computational cost, since they rely on sampling many rollouts for each prompt. Moreover, in RLVR the relative advantage is often sparse: many samples become nearly all-correct or all-incorrect, yielding low within-group reward variance and thus weak learning signals. In this paper, we introduce arrol (Accelerating RLVR via online Rollout Pruning), an online rollout pruning method that prunes rollouts during generation while explicitly steering the surviving ones more correctness-balanced to enhance learning signals. Specifically, arrol trains a lightweight quality head on-the-fly to predict the success probability of partial rollouts and uses it to make early pruning decisions. The learned quality head can further weigh candidates to improve inference accuracy during test-time scaling. To improve efficiency, we present a system design that prunes rollouts inside the inference engine and re-batches the remaining ones for log-probability computation and policy updates. Across GRPO and DAPO on Qwen-3 and LLaMA-3.2 models (1B-8B), arrol improves average accuracy by +2.30 to +2.99 while achieving up to 1.7x training speedup, and yielding up to +8.33 additional gains in average accuracy in test-time scaling. The code is available at https://github.com/Hsu1023/ARRoL.

cs.CL↗
arXiv2026

AutoCSF: Provably Space-Efficient Indexing of Skewed Key-Value Workloads via Filter-Augmented Compressed Static Functions

We study the problem of building space-efficient, in-memory indexes for massive key-value datasets with highly skewed value distributions. This challenge arises in many data-intensive domains and is particularly acute in computational genomics, where $k$-mer count tables can contain billions of entries dominated by a single frequent value. While recent work has proposed to address this problem by augmenting compressed static functions (CSFs) with pre-filters, existing approaches rely on complex heuristics and lack formal guarantees. In this paper, we introduce a principled algorithm, called AutoCSF, for combining CSFs with pre-filtering to provably handle skewed distributions with near-optimal space usage. We improve upon prior CSF pre-filtering constructions by (1) deriving a mathematically rigorous decision criterion for when filter augmentation is beneficial; (2) presenting a general algorithmic framework for integrating CSFs with modern set membership data structures beyond the classic Bloom filter; and (3) establishing theoretical guarantees on the overall space usage of the resulting indexes. Our open-source implementation of AutoCSF demonstrates space savings over baseline methods while maintaining low query latency.

cs.DS↗
arXiv2026

Atomically-sharp magnetic soliton in the square-net lattice EuRhAl$_{4}$Si$_{2}$

Topological spin textures are hallmark manifestations of competing interactions in magnetic matter. Their effective description by nonlinear field theories reflects an energetic frustration that destabilizes uniform order while selecting finite-size, topologically nontrivial configurations as stationary states. Among the most extreme realizations are atomically-sharp domain wall excitations, namely one-dimensional (1D) magnetic solitons, which represent the ultimate scaling limit of magnetic textures. Such solitons may emerge in magnetic systems where effective exchange interactions compete directly with uniaxial magnetic anisotropy. Here we show that the square-net rare earth compound EuRhAl$_{4}$Si$_{2}$ realizes a very susceptible regime where the magnetic anisotropy competes with highly frustrated exchange interactions stabilizing a rare ferrimagnetic $\uparrow\uparrow\downarrow$ state that, under applied magnetic field, supports the formation of atomically-sharp soliton defects. We confirm the bulk response of the 1D magnetic solitons via magnetization and electrical transport measurements. We establish both the zero- and in-field $\uparrow\uparrow\downarrow$ order via neutron diffraction, while magnetic force microscopy visualizes its real-space evolution into a stripe-like array. To elucidate the microscopic origin of the soliton, we relate the Ruderman-Kittel-Kasuya-Yosida (RKKY)-driven exchange interactions and the magnetic anisotropy through density functional theory, and we construct an effective 1D $J_{1}$-$J_{2}$-$K$ model whose atomistic spin dynamics simulations reproduce the observed soliton states as a function of external field. Our results demonstrate that EuRhAl$_{4}$Si$_{2}$ hosts atomically-sharp, field-driven 1D magnetic solitons, providing a new platform for studying 1D topological excitations at the atomic length scale.

cond-mat.str-el↗
arXiv2026

How Far Are LLMs from Professional Poker Players? Revisiting Game-Theoretic Reasoning with Agentic Tool Use

As Large Language Models (LLMs) are increasingly applied in high-stakes domains, their ability to reason strategically under uncertainty becomes critical. Poker provides a rigorous testbed, requiring not only strong actions but also principled, game-theoretic reasoning. In this paper, we conduct a systematic study of LLMs in multiple realistic poker tasks, evaluating both gameplay outcomes and reasoning traces. Our analysis reveals LLMs fail to compete against traditional algorithms and identifies three recurring flaws: reliance on heuristics, factual misunderstandings, and a "knowing-doing" gap where actions diverge from reasoning. An initial attempt with behavior cloning and step-level reinforcement learning improves reasoning style but remains insufficient for accurate game-theoretic play. Motivated by these limitations, we propose ToolPoker, a tool-integrated reasoning framework that combines external solvers for GTO-consistent actions with more precise professional-style explanations. Experiments demonstrate that ToolPoker achieves state-of-the-art gameplay while producing reasoning traces that closely reflect game-theoretic principles.

cs.AI↗
arXiv2026

Intention Knowledge Graph Construction for User Intention Relation Modeling

Understanding user intentions is challenging for online platforms. Recent work on intention knowledge graphs addresses this but often lacks focus on connecting intentions, which is crucial for modeling user behavior and predicting future actions. This paper introduces a framework to automatically generate an intention knowledge graph, capturing connections between user intentions. Using the Amazon m2 dataset, we construct an intention graph with 351 million edges, demonstrating high plausibility and acceptance. Our model effectively predicts new session intentions and enhances product recommendations, outperforming previous state-of-the-art methods and showcasing the approach's practical utility.

cs.CL↗
arXiv2025

Towards Effective Model Editing for LLM Personalization

Personalization is becoming indispensable for LLMs to align with individual user preferences and needs. Yet current approaches are often computationally expensive, data-intensive, susceptible to catastrophic forgetting, and prone to performance degradation in multi-turn interactions or when handling implicit queries. To address these challenges, we conceptualize personalization as a model editing task and introduce Personalization Editing, a framework that applies localized edits guided by clustered preference representations. This design enables precise preference-aligned updates while preserving overall model capabilities. In addition, existing personalization benchmarks frequently rely on persona-based dialogs between LLMs rather than user-LLM interactions, or focus primarily on stylistic imitation while neglecting information-seeking tasks that require accurate recall of user-specific preferences. We introduce User Preference Question Answering (UPQA), a short-answer QA dataset constructed from in-situ user queries with varying levels of difficulty. Unlike prior benchmarks, UPQA directly evaluates a model's ability to recall and apply specific user preferences. Across experimental settings, Personalization Editing achieves higher editing accuracy and greater computational efficiency than fine-tuning, while outperforming prompting-based baselines in multi-turn conversations and implicit preference questions settings.

cs.CL↗