arXiv subjects

Yuqi Chen

Publications and source records attributed to Yuqi Chen.

At least 19 recordsLinked to original sources

arXiv2026

Tunable Tool-Call Rates in LLM Agents via Representation Steering

Deciding whether to call a tool is a core competence of an LLM agent, and a costly one to get wrong: needless calls add latency, accrue cost, and may trigger irreversible side effects, while missing calls leave the model confidently wrong on questions it could only answer through tool-calls. Models manage this balance poorly, both over-using and under-using tools. Existing methods such as post-training and prompt engineering are expensive and difficult to modify at inference time. We show that whether an instruction-tuned model calls a tool can be controlled by a single linear direction in its residual stream, extracted without any training from the model's own tool-use preference signal and turned into an inference-time intervention with no prompt change. Adding the direction with strength $\alpha$ moves the call rate monotonically from near $0\% $ to over $90\%$ while keeping calls well-formed. The steering works in both directions: dialing it down suppresses calls, and dialing it up induces new calls that land precisely on the questions the model cannot answer from its own knowledge. We also show that the direction generalizes to unseen tools with strength comparable to each tool's own direction and without favoring any specific tool choice. With live tool execution, a single sweep of the steering traces a cost/accuracy Pareto frontier and nearly doubles open-domain QA accuracy ($0.29 \! \rightarrow \! 0.56$); the same recipe transfers across a diverse range of models spanning dense, MoE, and multimodal architectures, without any training. Our code is publicly available at https://github.com/YuqiChen4188/Steering-Tool-Use-Propensity.

cs.AI↗

arXiv2026

When Stories Evolve: Benchmarking LLM Storytelling Across Agent Architectures in Open-Ended World Simulations

Large language models can write fluent stories, but open-ended storytelling requires more than local fluency. In evolving world simulations and AI-native games, models must preserve facts, relationships, causal dependencies, and character states as the world changes. We introduce WSE-bench, a process benchmark that separately evaluates sustained generation, canonical coherence, and meaningful development in dynamic LLM storytelling. Generation Coverage records the proportion of planned narrative steps produced; Consistency tracks when canon breaks; and Richness measures how meaningfully branching, player-shaped trajectories develop. Across frontier models, Consistency and Richness do not form a smooth trade-off: their empirical Pareto frontier is non-concave, with several non-dominated intermediate configurations that no positive linear weighting can select. Added structure can enrich trajectories, but it does not uniformly improve coherence and may shorten them. Model scale chiefly improves sustained generation, without producing reliable gains in canonical coherence or meaningful development. These results show that sustained generation, canonical coherence, and meaningful development are distinct and sometimes competing capacities. WSE-bench makes those dynamics visible by extending narrative evaluation from finished stories to the processes that create them.

cs.CL↗

arXiv2026

From Scalars to Time Series: Rethinking Implicit Neural Representations for Time-Varying Volumetric Data

Implicit neural representations (INRs) for time-varying volumetric data are typically trained using dense sampling over spatiotemporal coordinates, where each observation corresponds to a single point in space and time. This coordinate-wise formulation requires extensive sampling during optimization, leading to high computational cost and inefficient use of temporal structure. In this work, we revisit this design choice and show that dense spatiotemporal sampling is not necessary for learning time-varying fields. Instead, we represent the data as a collection of spatially indexed time series and train INRs using sequence-level supervision over each spatial location, rather than coordinate-wise scalar samples. This reformulation eliminates the need for dense spatiotemporal sampling and instead learns each spatial location from its full temporal evolution in a structured manner. We demonstrate that this representation is compatible with a range of existing INR architectures and consistently improves reconstruction quality, while significantly reducing training cost. Furthermore, we show that this formulation can be combined with mixture-of-experts architectures, and that our MoE instantiation further improves reconstruction quality compared to both the base reformulation and existing MoE-based INR methods, providing a stronger capacity allocation under heterogeneous temporal dynamics.

cs.AI↗

arXiv2026

Leveraging Raman response in X-cut thin-film lithium tantalate for ultrabroadband combs and polychromatic visible light

X-cut thin-film lithium tantalate (TFLT) offers a unique combination of third nonlinearity, electro-optic effects, and a high optical damage threshold. However, its strong Raman response has historically hindered broadband Kerr comb generation. Here, we leverage this inherent Raman response by engineering coupling-defined dissipation. This allows us to reconfigure the relative thresholds of Raman and Kerr processes without modifying the intrinsic microresonator dispersion. Through this coupling-engineered threshold control, we can deliberately access distinct comb states, ranging from pure Kerr combs to Raman-Kerr synergistic broadband combs. We demonstrate a Kerr comb spanning 450 nm and a Raman-Kerr comb spanning 650 nm, representing the broadest combs reported to date on X-cut TFLT platforms. Moreover, in strongly coupled devices, we show that a single near-infrared pump can generate visible emission across multiple bands (from violet to red) via cascaded second sum-frequency processes. Our work demonstrates that a strong Raman response can be transformed from a parasitic competitor into an enabling mechanism for achieving broader comb spectra and generating polychromatic visible light. This work establishes X-cut TFLT as a powerful monolithic platform for nonlinear light sources, electro-optic functions, and complex photonic systems.

physics.optics↗

arXiv2026

Controlling Tool Use with Heading-Specific Activation Steering

Tool-augmented large language models extend their capabilities beyond parametric knowledge through external tools, but tend to invoke them unnecessarily. We investigate whether tool-use decisions have any stable internal representation that can be extracted and manipulated, a question that is non-trivial given that tools exist entirely in context at inference time and have no direct encoding in model weights. We show that steering vectors extracted from heading-anchors positions exert bidirectional causal control over tool-invocation behavior across five open-source models and three domains, suppressing unnecessary tool use most effectively in domains where parametric reasoning suffices. However, geometric analysis reveals that this causal effectiveness does not correspond to clean linear structure: tool-invocation steps exhibit diffuse, bimodal alignment with the suppression vector rather than the consistent negative alignment a linear encoding account would predict, and different tool types recruit largely distinct internal signatures with low cross-tool feature overlap. We hypothesize these geometric properties are indicative of the non-parametric nature of tools, and distinguish tool-use steering vectors from those extracted for parametrically grounded concepts. The relationship between this geometric irregularity and the observed causal effectiveness remains an open question.

cs.AI↗

arXiv2026

Flying to Image-Specified Objects: 3D Quadrotor Navigation via Cross-Graph Memory and Viewpoint Planning

Instance-Specific Image-Goal Navigation (InstanceImageNav) requires a robot to navigate toward the exact object instance depicted in a query image. Extending this task to quadrotors is challenging due to continuous 3D control, limited field of view (FOV), and safety constraints, which make successful navigation highly dependent on selecting informative viewpoints. We propose a hierarchical navigation framework for quadrotor InstanceImageNav that separates high-level decision making from low-level motion execution. Instead of navigating directly to spatial locations, the system generates viewpoint-aware action nodes around frontier regions and potential target objects, enabling the robot to explore while maintaining informative viewpoints for detecting the target instance. A lightweight semantic memory maintains object-level and observation-level context, allowing semantic cues to propagate to candidate action nodes for decision making. A learning-based policy selects the most promising action node, and a trajectory planner generates dynamically feasible 3D flight paths for safe execution. Experiments in simulation demonstrate consistent improvements over strong baselines, and real-world quadrotor flights validate the practicality and robustness of the proposed framework.

cs.RO↗

arXiv2026

RoamFlow: Reinforcement-Aligned One-Step Action MeanFlow Policy for Image-Goal Navigation

Image-goal navigation is a key challenge in embodied robotics, where an agent must reach a target specified solely by a goal image. While existing reinforcement learning approaches map perceptual observations directly to actions, they struggle to model long-horizon dependencies, often leading to suboptimal trajectories. To address this limitation, we propose RoamFlow, a generative navigation framework that leverages MeanFlow to predict the average velocity field for trajectory synthesis, enabling efficient few-step generation and reducing inference latency. We further adopt a two-stage training strategy that combines expert imitation for stable initialization with reinforcement learning for task-specific policy refinement. Extensive experiments in both Habitat simulation and real-world robotic platforms demonstrate that RoamFlow achieves efficient inference while maintaining strong navigation performance under real-time constraints.

cs.RO↗

arXiv2026

A Gradient Perspective on RLVR Stability and Winner Advantage Policy Optimization

Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning, but GRPO-style optimization remains prone to collapse. We analyse this instability through token-level gradient dynamics, deriving a taxonomy that predicts how updates affect next-token probabilities and entropy. The taxonomy shows that stability depends jointly on the advantage sign and token distribution under the current policy. Motivated by this finding, we propose Winner Advantage Policy Optimization (WAPO), a simple online clipped policy-gradient objective that updates only on positive-advantage completions. Across mathematical reasoning and multi-hop QA benchmarks, WAPO improves training stability and matches or outperforms baselines across multiple model families. Full code can be found at https://github.com/layer6ai-labs/wapo.

cs.LG↗

arXiv2026

AgentBeats: Agentifying Agent Assessment for Openness, Standardization, and Reproducibility

Agent systems are advancing quickly across domains, but their evaluation remains fragmented. Most benchmarks rely on fixed, LLM-centric harnesses that require heavy integration, create test-production mismatch, and limit fair comparison across diverse agent designs. The root problem is the lack of an open, agent-agnostic assessment interface. We advocate Agentified Agent Assessment (AAA), where evaluation is performed by judge agents and all participants interact through standardized protocols: A2A for task management and MCP for tool access. Conventional benchmarking defines two separate interfaces, one for the benchmark and one for the agent, while AAA only needs one; this yields a generic, unified framework that separates assessment logic from agent implementation and enables reproducible, interoperable, and multi-agent evaluation. We further introduce AgentBeats as a concrete realization of AAA: we identify five practical operation modes that make standardized assessment compatible with real-world constraints on openness, privacy, and reproducibility. To evaluate our design at scale, we conduct two studies: a five-month open competition that drew 298 judge agents across 12 categories together with 467 subject agents from independent participants, showing that AAA applies across a heterogeneous range of benchmarks; and a case study on coding agents that confirms agentified evaluation preserves fidelity with the public record while surfacing previously missing head-to-head results, yielding research insights about agent design. Combining a community-scale field study and a controlled coding case study, we verify that AAA delivers coverage, practicality, and fidelity across heterogeneous scenarios at scale. Together, AAA and AgentBeats offer a clear path toward open, standardized, and reproducible agent assessment.

cs.AI↗

arXiv2026

Divination by Prompt: LLM-Mediated Xuanxue on Chinese Social Media

The rapid proliferation of large language models (LLMs) has produced a striking cultural practice: using conversational AI for divination. This paper offers one of the first systematic studies of LLM-mediated divination in the context of Xuanxue, an internet-native umbrella term for mystical and spiritual practices on Chinese social media. Using a mixed-methods design, we analyze 23000+ posts and comments from Xiaohongshu and conduct 32 semi-structured interviews with users and professional diviners. Users primarily consult LLMs about pragmatic concerns - romantic relationships, careers, exams, and in-game gacha draws - via two intersecting pathways: trend-driven curiosity enabled by viral visibility and zero-cost access, and event-driven anxiety under conditions of uncertainty. A defining feature is collaborative prompt refinement, which turns users into active prompt engineers. Among commenters expressing a clear stance, perceived efficacy skews positive, with "accuracy" often justified through biographical fit and retrospective confirmation, consistent with Barnum and confirmation bias. Users also develop verification practices such as repeated trials and cross-model comparison. Professional diviners, by contrast, portray LLMs as lacking the "spiritual power" required for genuine divination, reflecting both ontological commitments and economic boundary-work. We also show how participants navigate tensions between scientific and metaphysical frames when interpreting AI-generated readings. Situating these findings in anthropological and cognitive-evolutionary theories of divination, we argue that LLM divination preserves core functions of traditional practice while introducing scalability, repeatability, and prompt-driven co-production that reshape how divinatory authority is constructed and evaluated.

cs.CY↗

arXiv2026

Efficient Community Search on Attributed Public-Private Graphs

Public-private graph, where a public network is visible to everyone and every user is also associated with its own small private graph accessed by itself only, widely exists in real-world applications of social networks and financial networks. Most existing work on community search, finding a query-dependent community containing a given query, only studies on a public graph, neglecting the privacy issues in public-private networks. However, considering both the public and private attributes of users enables community search to be more accurate, comprehensive, and personalized to discover hidden patterns. In this paper, we study a novel problem of attributed community search in public-private graphs (ACS-PP), aiming to find a connected k-core community that shares the most keywords with the query node. This problem uncovers structurally cohesive communities, such as interest-based user groups or core teams in collaborative networks. To optimize search efficiency, we propose an integrated scheme of constructing a public global graph index and a private personalized graph index. For the private index, we developed a compact structure of the PP-FP-tree index. The PP-FP-tree is constructed based on the public and private neighbors of the query node in the public-private graph, serving as an efficient index to mine frequent node sets that share the most common attributes with the query node. Extensive experiments on real public-private graph datasets validate both the efficiency and quality of our proposed PP-FP search algorithm against existing competitors. The case study on public-private collaboration networks provides insights into the discovery of public-private communities.

cs.DB↗

arXiv2026

High-power-handling ultra-compact acousto-optic modulators using one-dimensional topological interface states on thin-film lithium tantalate

Recent advances in integrated photonics have enabled on-chip signal modulation and processing through localized photon-phonon interactions. For acousto-optic devices, compact footprint and high efficiency are essential for dense integration, while strong power handling is critical for stable operation in demanding applications. However, it remains challenging to achieve these features simultaneously on existing integrated platforms. Here, we propose and experimentally demonstrate, for the first time on a thin-film lithium tantalate platform, an ultra-compact acousto-optic modulator based on topological interface states. Benefiting from the strong optical confinement of the topological boundary state, the device achieves a footprint of 0.13 by 0.12 mm2 and a half-wave voltage-length product of 0.491 Vcm. We further demonstrate stable acousto-optic modulation at an on-chip optical power of up to 28 dBm (630.9 mW), highlighting the strong power-handling capability of the thin-film lithium tantalate topological structure. This work provides a compact and high-power solution for microwave-to-photonic transduction and shows the potential of the thin-film lithium tantalate for robust integrated photonic systems.

physics.optics↗

arXiv2026

Turning Generators into Retrievers: Unlocking MLLMs for Natural Language-Guided Geo-Localization

Natural-language Guided Cross-view Geo-localization (NGCG) aims to retrieve geo-tagged satellite imagery using textual descriptions of ground scenes. While recent NGCG methods commonly rely on CLIP-style dual-encoder architectures, they often suffer from weak cross-modal generalization and require complex architectural designs. In contrast, Multimodal Large Language Models (MLLMs) offer powerful semantic reasoning capabilities but are not directly optimized for retrieval tasks. In this work, we present a simple yet effective framework to adapt MLLMs for NGCG via parameter-efficient finetuning. Our approach optimizes latent representations within the MLLM while preserving its pretrained multimodal knowledge, enabling strong cross-modal alignment without redesigning model architectures. Through systematic analysis of diverse variables, from model backbone to feature aggregation, we provide practical and generalizable insights for leveraging MLLMs in NGCG. Our method achieves SOTA on GeoText-1652 with a 12.2% improvement in Text-to-Image Recall@1 and secures top performance in 5 out of 12 subtasks on CVG-Text, all while surpassing baselines with far fewer trainable parameters. These results position MLLMs as a robust foundation for semantic cross-view retrieval and pave the way for MLLM-based NGCG to be adopted as a scalable, powerful alternative to traditional dual-encoder designs. Project page and code are available at https://yuqichen888.github.io/NGCG-MLLMs-web/.

cs.CV↗

arXiv2026

Optimal Circuit Synthesis of Linear Codes for Error Detection and Correction

Fault injection attacks deliberately inject faults into a device via physical channels to disturb its regular execution. Adversaries can effectively deduce secrets by analyzing both the normal and faulty outputs, posing serious threats to cryptographic primitives implemented in hardware. An effective countermeasure to such attacks is via redundancy, commonly referred to as concurrent error detection schemes, where Binary linear codes have been used to defend against fault injection attacks. However, designing an optimal code circuit is often time-consuming, error-prone, and requires substantial expertise. In this paper, we formalize the optimal code circuit synthesis problem (OptiCC) based on two domain-specific minimization objectives on individual inputs and parity size. We then propose a novel algorithm CiSC for solving OptiCC, prioritizing the minimization of individual inputs. Our approach features both correct-by-construction and secure-by-construction. In a nutshell, CiSC gradually reduces individual inputs and parity size by checking, via SMT solving, the existence of feasible Boolean functions for implementing a desired code. We further present an effective technique to lazily generate combinations of inputs to Boolean functions, while quickly identify equivalent ones. We implement our approach in a tool CiSC, and evaluate it on practical benchmarks. Experimental results show our approach can synthesize code circuits that significantly outperform those generated by the latest state-of-the-art techniques.

cs.CR↗

arXiv2025

Causal-Guided Detoxify Backdoor Attack of Open-Weight LoRA Models

Low-Rank Adaptation (LoRA) has emerged as an efficient method for fine-tuning large language models (LLMs) and is widely adopted within the open-source community. However, the decentralized dissemination of LoRA adapters through platforms such as Hugging Face introduces novel security vulnerabilities: malicious adapters can be easily distributed and evade conventional oversight mechanisms. Despite these risks, backdoor attacks targeting LoRA-based fine-tuning remain relatively underexplored. Existing backdoor attack strategies are ill-suited to this setting, as they often rely on inaccessible training data, fail to account for the structural properties unique to LoRA, or suffer from high false trigger rates (FTR), thereby compromising their stealth. To address these challenges, we propose Causal-Guided Detoxify Backdoor Attack (CBA), a novel backdoor attack framework specifically designed for open-weight LoRA models. CBA operates without access to original training data and achieves high stealth through two key innovations: (1) a coverage-guided data generation pipeline that synthesizes task-aligned inputs via behavioral exploration, and (2) a causal-guided detoxification strategy that merges poisoned and clean adapters by preserving task-critical neurons. Unlike prior approaches, CBA enables post-training control over attack intensity through causal influence-based weight allocation, eliminating the need for repeated retraining. Evaluated across six LoRA models, CBA achieves high attack success rates while reducing FTR by 50-70\% compared to baseline methods. Furthermore, it demonstrates enhanced resistance to state-of-the-art backdoor defenses, highlighting its stealth and robustness.

cs.CR↗

arXiv2025

Developing a Strong CPS Defender: An Evolutionary Approach

Cyber-physical systems (CPSs) are used extensively in critical infrastructure, underscoring the need for anomaly detection systems that are able to catch even the most motivated attackers. Traditional anomaly detection techniques typically do `one-off' training on datasets crafted by experts or generated by fuzzers, potentially limiting their ability to generalize to unseen and more subtle attack strategies. Stopping at this point misses a key opportunity: a defender can actively challenge the attacker to find more nuanced attacks, which in turn can lead to more effective detection capabilities. Building on this concept, we propose Evo-Defender, an evolutionary framework that iteratively strengthens CPS defenses through a dynamic attacker-defender interaction. Evo-Defender includes a smart attacker that employs guided fuzzing to explore diverse, non-redundant attack strategies, while the self-evolving defender uses incremental learning to adapt to new attack patterns. We implement Evo-Defender on two realistic CPS testbeds: the Tennessee Eastman process and a Robotic Arm Assembly Workstation, injecting over 600 attack scenarios. In end-to-end attack detection experiments, Evo-Defender achieves up to 2.7% higher performance than state-of-the-art baselines on unseen scenarios, while utilizing training data more efficiently for faster and more robust detection.

cs.CR↗

arXiv2025

VMDT: Decoding the Trustworthiness of Video Foundation Models

As foundation models become more sophisticated, ensuring their trustworthiness becomes increasingly critical; yet, unlike text and image, the video modality still lacks comprehensive trustworthiness benchmarks. We introduce VMDT (Video-Modal DecodingTrust), the first unified platform for evaluating text-to-video (T2V) and video-to-text (V2T) models across five key trustworthiness dimensions: safety, hallucination, fairness, privacy, and adversarial robustness. Through our extensive evaluation of 7 T2V models and 19 V2T models using VMDT, we uncover several significant insights. For instance, all open-source T2V models evaluated fail to recognize harmful queries and often generate harmful videos, while exhibiting higher levels of unfairness compared to image modality models. In V2T models, unfairness and privacy risks rise with scale, whereas hallucination and adversarial robustness improve -- though overall performance remains low. Uniquely, safety shows no correlation with model size, implying that factors other than scale govern current safety levels. Our findings highlight the urgent need for developing more robust and trustworthy video foundation models, and VMDT provides a systematic framework for measuring and tracking progress toward this goal. The code is available at https://sunblaze-ucb.github.io/VMDT-page/.

cs.CV↗

arXiv2025

Ultrafast Reconfigurable Topological Photonic Processing Accelerator

The rise of artificial intelligence has triggered exponential growth in data volume, demanding rapid and efficient processing. High-speed, energy-efficient, and parallel-scalable computing hardware is thus increasingly critical. We demonstrate a wafer-scale non-volatile topological photonic computing chip using topological modulators. Leveraging the GHz-speed electro-optic response and nonvolatility of ferroelectric lead zirconate titanate (PZT) thin films via topological photonic confinement, Our chip enables thousand-fold faster reconfiguration, zero-static-power operation, and a computational density of 266 trillion operations per second per square millimeter . This density surpasses that of silicon photonic reconfigurable computing chips by two orders of magnitude and thin-film lithium niobate platforms by four orders of magnitude. A 16-channel wavelength-space multiplexed chip delivers 1.92 TOPS throughput with 95.64% digit-recognition accuracy and 94.5% precision for solving time-varying partial differential equations. Additionally, the chip supports functional reconfiguration for high bandwidth density optical I/O. This work establishes ferroelectric topological photonics for efficient high-speed photonic tensor processing.

physics.optics↗