Papers

As LLM-generated content becomes more sophisticated, detection systems for distinguishing those texts from human-written text must operate at scale while handling diverse writing styles, domains, languages, and generator models. Input-only encoder detectors are suitable for practical deployment setting, but standard binary classification supplies only the class label and does not explicitly organize the substantial variation within either class. We propose MD-ProTector, which represents each class with multiple trainable reference vectors in the encoder embedding space, referred to as prototypes. These prototypes provide separate decision boundaries for different groups of texts within the same class. However, adding multiple prototypes alone does not determine which variation each prototype should represent. MD-ProTector addresses this problem with Prototype Positioning loss, which separates class-level structure from the within-class variation that differentiates individual prototypes. Evaluated across five settings from three large-scale benchmarks covering domain, generator, language, and adversarial variation, MD-ProTector achieves the highest AvgRec on MAGE CDCM and RAID and the highest AUROC and lowest FPR95 on RAID among the compared encoder-based methods.

Complex software systems develop over timescales that exceed the lifespan of any individual coding agent. Most agentic software systems preserve continuity through persistent sessions, memories, managers or shared context. We introduce EvoX Genesis (hereafter, Genesis), which instead makes the software project persistent while allowing local agents to remain finite-lived. Genesis represents software as a persistent recursive world: each local world is situated by an accepted version and a repository path, finite-lived agents propose local changes, recursive delegation moves work across paths, and only accepted consequences advance the persistent version history. We evaluate this organization across formation, continuation and redevelopment. Starting from a repository with no compiler implementation, Genesis used DeepSeek V4 Flash to build a Rust-based C compiler with about 250k tracked lines; the run lasted over 120 hours, archived over 1,000 agent episodes and incurred only US\$44 in model-token charges. The compiler passed the complete c-testsuite and most LLVM and Csmith tests. In a separate compiler world generated with GLM 5.2, development continued after repeated agent replacement while retaining full test performance. Genesis also reimplemented 13 MESA modules with over 100k Fortran lines as a Rust workspace with nearly 90k Rust lines; across six numerical workloads, it achieved median speedups of 1.55–6.87x. These results show that long-horizon software development can be organized around a persistent project rather than a persistent agent. Project Website: https://genesis.evox.group/

Multimodal emotion recognition in conversation (MERC) requires understanding complex interactions between verbal and non-verbal cues. However, most existing approaches fundamentally treat this as a direct input-output (multimodal cues-emotion labels) mapping problem, overlooking the causal reasoning that humans use when interpreting emotions. We propose rationale-guided learning (RGL), a novel framework that transforms MERC into a cognitively-inspired reasoning task. Based on dual-process theory, we decompose emotional reasoning into three facets: Intuitive (immediate perception, System 1), Contextual (situational analysis, System 2), and Integrative (synthesis of both). We leverage an MLLM offline to generate structured rationales, which are encoded as memories to guide model training via aligning internal representations with human-like reasoning patterns. Our final model operates without any MLLM overheads at inference time. Experimental results show that RGL achieves state-of-the-art performance on the IEMOCAP and MELD benchmarks. Further, for interpretation, we demonstrate that the model's internal features effectively retrieve semantically correct rationales for unseen test samples, validating its rationale reasoning capabilities.

Large language model-based recommender systems are increasingly adopting slow-thinking models that generate step-by-step reasoning before making predictions, often achieving higher accuracy than fast-thinking models that predict directly. However, their reasoning traces are often unnecessarily verbose, increasing inference costs without commensurate accuracy gains. Existing training-based approaches to reasoning compression often incur substantial adaptation costs, while inference-time methods are brittle and difficult to scale. These limitations motivate model merging as a promising training-free direction for transferring specialised behaviours between models in a shared parameter space. In particular, merging a slow-thinking model with a fast-thinking counterpart provides a natural mechanism for balancing recommendation accuracy and reasoning conciseness. To this end, we propose, to our knowledge, the first model merging framework for reasoning compression in recommender systems. Unlike conventional merging methods that apply uniform merge coefficients across model components, our method performs fine-grained merging at the level of individual attention heads, capturing heterogeneous patterns in recommendation reasoning. Each attention head is assigned a distinct merge coefficient according to its contribution to critical reasoning evidence and its sensitivity to parameter change, enabling selective injection of the concise behaviour of the fast-thinking model into the slow-thinking model and reducing reasoning verbosity without compromising recommendation quality. Experiments on three benchmark datasets show that our method reduces reasoning length by up to 24.3% while outperforming competitive model merging baselines in maintaining recommendation accuracy. The code is available at https://github.com/linhledieu/REAM.

Large language models (LLMs) have made substantial progress on reasoning tasks that require increasingly long and complex inferential chains. This progress primarily reflects reasoning depth. A complementary and comparatively unexamined capability is reasoning breadth: exploring multiple semantic directions in parallel and integrating the resulting clues into one coherent answer. We introduce MPAR-Bench, a bilingual English-Chinese benchmark that isolates reasoning breadth through multi-point associative reasoning. Inspired by the cooperative game Just One, each item asks a model to recover a hidden target from several independently generated, semantically diverse clues. We construct 1,000 items using a multi-agent clue-generation pipeline, embedding-based diversity filtering, and human verification. Only the answer space is drawn from public word lists, whereas every clue set is generated from scratch. Beyond exact-match accuracy, we evaluate models using accuracy, ANLS, embedding similarity, reasoning-trace verification, and four perturbations: clue masking, order shuffling, distractor injection, and multi-step clues. Across evaluated models, perturbations reduce accuracy by 9-18 percentage points in English and 5-12 percentage points in Chinese. Thinking mode improves standard-setting accuracy, especially in English, but does not consistently reduce sensitivity to perturbations. Case-level analysis also shows that extended reasoning can overturn an initially correct hypothesis. These results indicate that greater reasoning depth does not automatically confer robust reasoning breadth, and that reasoning breadth remains largely uncovered by current benchmarks.

Automatic stress detection from facial video offers a practical path to non-intrusive affect monitoring, yet existing video-based approaches commonly decompose full recordings into short temporal windows before classification. This design introduces additional choices regarding window length, overlap, and aggregation, while limiting direct analysis of temporal information across the entire recording. In this study, we present FUSE (Frame-Unified Stress Estimation), a facial-video stress detection framework that processes complete recordings as a single input without temporal windowing or external segmentation. The name reflects the defining operation of the method: rather than dividing a recording into short clips, all frames are fused into one unified two-dimensional representation from which the stress state is estimated. This unification is realized by folding the temporal dimension into the channel dimension of the spatial representation, and the resulting high-dimensional input is processed using a unified asymmetric-attention architecture. At a temporal stride of t = 1, FUSE retains the full 120-second recording as one input, corresponding to 3,600 frames at 30 fps. Experiments on a 58-subject stress dataset using a stratified subject-level protocol evaluate seven temporal-stride configurations, ranging from full-frame input to sparse subsampling. FUSE achieves the highest test accuracy of 69.44% at t = 15, while the full-frame configuration remains competitive at 69.03%. Across the stride range, computational cost varies from 12.48 to 348.78 GFLOPs, showing the trade-off between temporal density and efficiency. These results demonstrate that temporal windowing is not required for effective facial-video stress detection in this setting, and that complete-recording inference can be achieved within a single unified architecture.

Large language models increasingly rely on external tools to access up-to-date information, perform computation, and interact with the outside world. For autoregressive models, tool use naturally fits the generation process: the model emits a tool call, waits for the result, and then continues generating. Diffusion language models (dLLMs), however, reason by repeatedly refining many parts of their output in parallel, making this stop-and-resume interaction pattern unnecessarily restrictive. It can force tool decisions before the model's reasoning has stabilized, delay useful observations until a discrete call finishes, and introduce redundant refinement and tool execution, potentially hurting both task accuracy and inference efficiency. We introduce Continuous Interaction Diffusion (CID), a diffusion-native model–runtime architecture that integrates tool interaction into iterative denoising. CID separates a model-read-only fact channel, a thought channel represented by a Typed Cognitive Tensor, and a display channel. Information needs can emerge before a textual or JSON call is fully serialized, allowing perceptual bindings to launch external reads while denoising continues. Returned results are projected into the evolving thought state and can revise earlier cognition and display regions. Persistent bindings reuse static results without repeated external execution and refresh changing sources when needed. CID is designed to expose evidence earlier, overlap tool latency with model computation, reduce duplicate external work, and preserve useful computation after new evidence arrives. We formalize the architecture, runtime, and training objectives, and define an evaluation protocol for task quality and end-to-end efficiency. This first paper focuses on read-only tools and makes no empirical performance claims.

Machine learning-based Intrusion Detection Systems (IDS) have demonstrated superior performance in securing Unmanned Aerial Vehicle (UAV) networks. However, the 'black-box' nature of these models, combined with the high dimensionality of multimodal cyber-physical data, poses significant interpretability challenges. Static visualization dashboards may struggle to present complex relationships among multimodal cyber-physical features in a form that is easy for operators to inspect and interpret. To address this, we propose a Conversational XAI interface powered by Large Language Models (LLM) to facilitate on-demand investigation. In a controlled experiment with participants, we systematically evaluated the impact of this conversational interface versus a traditional XAI Dashboard on operator understanding, trust, and reliance during post-incident auditing tasks. Our results suggest that the conversational interface was perceived as more useful than the dashboard, potentially because it helped participants access and synthesize relevant information more easily. However, this benefit was accompanied by a lower level of appropriate self-reliance, indicating a potential risk of over-reliance. One possible interpretation is that the natural-language responses made the AI advice easier to accept, which may have reduced participants' tendency to verify the underlying evidence when the IDS was incorrect. These findings point to a potential trade-off in human-AI collaboration for UAV intrusion auditing: interaction mechanisms that improve perceived usability may also increase the risk of inappropriate reliance. We conclude by discussing design implications for future XAI systems that balance seamless interaction with cognitive forcing functions to foster appropriate reliance.

Responsible AI (RAI) has become a central concern for technology companies, regulators, and the public. How industry practitioners interpret, implement, and sustain RAI work directly shapes the design and deployment of AI systems. As empirical scholarship examining RAI practices in industry has rapidly expanded, findings are dispersed across studies that focus on different roles, organizational contexts, and interventions. This work synthesizes current knowledge through a literature review of 161 empirical studies spanning six years, each engaging industry practitioners via interviews, surveys, workshops, ethnographies, and other methods. Our synthesis reveals both meaningful progress and persistent challenges in industry RAI practice. Practitioner awareness has increased, RAI activities have become more professionalized, and interventions such as toolkits and guidelines are more widely adopted. At the same time, practitioners continue to face substantial barriers, including limited training, uneven organizational support, and a lack of interventions tailored to day-to-day work practices. By consolidating and organizing these findings, we provide a more complete account of industry RAI than any single study to date. We conclude by discussing implications for RAI researchers, practitioners seeking to adopt effective practices, and policymakers aiming to ground governance efforts in the realities of industry contexts.

Large Language Models (LLMs) deployed as AI agents frequently exhibit user specification-grounding failures, executing hallucinated, undesired actions to force a resolution rather than expressing uncertainty. Existing detection methods fail to provide actionable, real-time correction as they either do not localize the hallucinations, or incur prohibitive inference latency. We introduce the Latent Critic, a lightweight low-rank adapter (LoRA) that operates concurrently with a frozen base LLM's generation to actively restructure the transformer's residual stream—amplifying latent grounding signals and translating them into localized, natural language feedback within a single sequence. By refining the base model's native uncertainty signals, this manipulation of the latent space enables reliable, granular detection without the overhead of secondary inference loops. Mechanistic analysis via activation patching and layer-wise probing shows that this rank-invariant behavior restructures pre-existing uncertainty geometry into a linearly separable representation that transfers more reliably than base model representations alone. Using tool-calling as an instantiation of granular hallucinations, we validate the detection and downstream improvements enabled by the Latent Critic architecture across Qwen and Llama-based models. Demonstrating superior real-time efficacy, our approach significantly outperforms equivalent-scale fine-tuned external detectors, semantic entropy baselines, and passive internal probes in isolating hallucinations, achieving 0.966 AUROC and >80% accuracy in localization (e.g., ungrounded: date). When deployed in a closed-loop ReAct environment, the Critic acts as a negligible latency guardrail, intercepting hallucinations before execution to prevent undesired actions while simultaneously leveraging this specific localized feedback to enable efficient agent self-correction.

Scattering matrices are the standard experimental and computational description of photonic and electromagnetic devices. Passivity is explicit in the conventional incoming-outgoing matrix, whereas causality sum rules are usually formulated only after transforming the response into auxiliary variables. Here we show that these rules can be written directly in the conventional scattering matrix by removing the time advance introduced by the reference domain. Using the earliest-arrival delay of each channel, we define a domain-delayed matrix that preserves real-frequency passivity while restoring the causal time origin. Under explicit analyticity, transparency, and regularity assumptions, this matrix becomes a Schur function, enabling a Cayley-Herglotz construction. The resulting projected and determinant bounds constrain coherent channel superpositions and aggregate multichannel loss. The framework recovers Rozanov's absorber limit and spherical-multipole sum rules, while extending causality bounds to measurable quantities including insertion loss, suppressed singular-value channels, and conditional lossless delay-bandwidth trade-offs. Our work directly connects fundamental causality theory with experimentally accessible scattering data. The initial theoretical route is autonomously explored by Qiushi Engine, an AI research system for open-ended scientific discovery, and subsequently verified, refined, and developed by the authors, demonstrating a hybrid AI-human discovery workflow.

A slew of recent works develop agents for solving research problems end-to-end, a paradigm increasingly referred to as autoresearch. Such agents have inspired large industry investment, motivated by their potential to automate time-consuming human labor and customize machine learning solutions for specialized applications. In this paper, we study the modeling pipeline at the core of these autoresearch systems and identify common failure modes when they are applied to tabular datasets: (1) they waste compute resolving the same bugs over and over again; (2) they often fail to tune hyperparameters even when they have a large remaining compute budget; (3) the tree-search algorithms that power them do not explore; and (4) they perform data analysis, mimicking the humans whose data they are trained on, but do not use that analysis to make downstream decisions. We explore targeted interventions and find that a global debug consultant that shares discovered runtime constraints across all branches of the search tree, prompt- and control-level enhancements, and refined tree-search algorithms successfully recover wasted compute. Our results show that large gains in autoresearch agent performance are achievable through agentic design alone, holding the underlying language model fixed.

Reasoning shortcuts are rule solutions that reach correct predictions through unintended concepts. A recent framework of Takemura, Inoue, and Nishino analyzes them through an automorphism group of value relabelings, asking when rules pin concepts down. Its key definition, one value permutation shared across all positions, does not apply as stated to any of its four heterogeneous benchmarks, and the most direct embedding, padding, produces confident false pathology: 90.91% of solution pairs unexplained on CLE4EVR, versus 0% under every well-defined rung of the componentwise hierarchy we introduce; the padded verdict rotates under configuration-file ordering. Across eleven rule families under fifteen pre-specified predictions (thirteen confirmed), unexplained-pair rates span 0% to 99.9999% and track provable structure: six theorems give sufficient conditions for transitivity and its failure. For circuit-given rules, symmetry-inertness of a coordinate is coNP-complete; automorphism existence is coNP-hard under randomized reductions, lies in $\Sigma_2^p$, is not $\Sigma_2^p$-complete in the Boolean case unless PH collapses, and is coNP-complete on monotone circuits. Boolean transitivity is classified exactly: automorphisms explain everything iff the solution set is an affine coset. Weakly supervised models place all 94 observed shortcuts at the one level the theory flags, none at the 48 it certifies transitive, and none at twelve typed-ambiguous levels. Relocating the absorbing element moves every shortcut with it; a confusion null attributes the location to geometry while the observed rate exceeds it by half again. Trained end to end on CLE4EVR's rule and heterogeneous domains through a synthetic prototype front end, models produce 20,223 label-preserving errors with zero different-orbit exceptions, as transitivity predicts, where the padded instrument would misreport 78-88% of them.

We present a theoretical foundation for inverse-distance attention, from its Euclidean prototype (Resolver) to its non-Euclidean realization (Riemann GeoResolver). The Euclidean part establishes three core theorems: (1) circuit separation—IDA achieves exact retrieval with $\mathcal{O}(1)$ resources while softmax requires $\Omega((\log n)^2)$ width; (2) a Polyak–Lojasiewicz inequality with $\Omega(e^{\Delta^2/\sqrt{d}}/\Delta^2)$ stronger constant than softmax, implying linear convergence, $\mathcal{O}(\log n)$ Lipschitz scaling under a low-rank/clustering assumption, $\Theta(1)$ Hessian spread, and absence of spurious local minima; (3) a width-independent effective rank bound that limits noise memorization—softmax memorizes arbitrary labels when $d_h\ge n$, while IDA limits test error to $\mathcal{O}(\eta^2)$. The non-Euclidean extension then builds upon this prototype, replacing Euclidean distance with hyperbolic geodesic distance for storage and spherical geodesic distance for routing. The Riemann GeoResolver framework comprises ten integrated modules: four HIDA operators spanning $\Theta(n^2)$ to $\Theta(1)$ per token; Hyperbolic Curvature Compression (HCC) with provable error bounds; HyperGate with gradient lower-bound theorem; Spherical Inverse Distance Attention (SIDA) with sphere-analog PL inequalities; Dynamic Memory Genesis (DMG) with $\mathcal{O}(\log T)$ regret bounds; and Geodesic Sparse Routing (GSR) with quality and communication bounds. The Euclidean theorems are proved in full; the non-Euclidean extension theorems are proved with analogous arguments. This work establishes a theoretical arc: from Euclidean attention as a special case, to hyperbolic memory, to spherical retrieval.

Many studies have shown that specially crafted inputs can induce large language models (LLMs) to generate excessively long outputs, resulting in significant computational overhead and resource consumption. While most existing denial-of-service (DoS) attacks target text-only LLMs, end-to-end (E2E) speech LLMs are rapidly emerging. Existing text-based DoS attacks primarily rely on prompt engineering, such as adversarial suffixes or semantic inducement, which exploit the discrete nature of text inputs and therefore cannot be directly transferred to continuous speech inputs. Moreover, prior studies on speech model security mainly focus on ASR or TTS systems, leaving the DoS vulnerability of E2E speech LLMs largely unexplored. To address this gap, we propose the perturbation-based DoS attack targeting E2E speech models. Instead of inducing long outputs through prompt manipulation, our method optimizes imperceptible acoustic perturbations to directly influence the model's autoregressive generation process while preserving the original input length. Specifically, we formulate the attack as a composite optimization objective that jointly suppresses EOS generation, encourages prolonged decoding, and largely preserves semantic consistency by integrating weighted EOS loss, top-k logit loss, length loss, and semantic alignment loss. To further improve stealthiness, we employ voice activity detection (VAD) to inject perturbations only into voiced regions. Extensive experiments on three open-source E2E speech LLMs demonstrate that our method achieves stable attack success rate while significantly increasing generation length and GPU resource consumption, revealing security risks in modern ALLMs.

Reinforcement learning (RL) has shown promising performance in autonomous driving, yet ensuring the safety of online RL policies remains challenging due to insufficient exposure to safety-critical driving scenes. The long-tailed nature of real-world traffic situations makes dangerous and rare interactions difficult to encounter through conventional sampling, limiting the ability of RL policies to learn robust safety behaviors. Existing methods improve training diversity by synthesizing challenging scenes or adversarial situations. However, these approaches typically optimize scene generation objectives separately from the evolving policy, without explicitly modeling how generated perturbations relate to the current policy's weaknesses and learning needs. In this paper, we propose Threat-guided Policy-aware Scene Perturbation (TPSP) for safe autonomous driving with online RL. TPSP introduces a policy-aware scene encoder to capture the interaction between policy behaviors and surrounding environments, enabling scene perturbation aligned with the current policy. Based on this representation, TPSP selectively perturbs critical objects rather than applying uniform modifications across the scene. Furthermore, we develop a threat-guided optimization strategy that evaluates perturbed scenes through threat-level differences between policy rollouts on original and perturbed scenes, guiding the generation of safety-critical scenes with higher training value. Comprehensive experiments demonstrate that TPSP improves safety learning efficiency, achieving strong safety performance on NAVSIM v2 with approximately 4 million kilometers of simulated driving data. Ablation studies verify that policy-aware targeted perturbations provide more informative safety-critical experiences than random or policy-unaware strategies, enabling safer driving under limited interaction budgets.

ELVAE places an input-dependent normal–inverse-gamma (NIG) hierarchy at each VAE latent coordinate, separating location uncertainty $u_{\mathrm{epi}}=\beta/[\nu(\alpha-1)]$ from conditional variability $u_{\mathrm{var}}=\beta/(\alpha-1)$. The marginalized latent law, however, identifies only the three quotient coordinates $(\gamma,\alpha,c)$ with $c=\beta(1+1/\nu)$; reconstruction is blind to one $(\nu,\beta)$ fiber direction. A companion theoretical analysis shows that the complete NIG prior and forward KL select a unique prior-relative canonical representative on each fiber, so canonical inverse allocation is not a fourth independent information channel. Empirically, trained inverse evidence $1/\nu$ remains the strongest sensitivity-ranking score. At $\tau_{\mathrm{epi}}=1$, the three-seed mean high/low-$u_{\mathrm{epi}}$ semantic-transition ratios are 1.98 on MNIST and 1.66 on Fashion-MNIST, falling to 1.33 and 1.16 under scale-matched controls. In the 20-draw MNIST component study with equal per-anchor perturbation energy, $1/\nu$ gives high/low ratios 1.69 and 1.65 under the $u_{\mathrm{epi}}$ and $u_{\mathrm{var}}$ fields and 1.69 (95% interval 1.45–1.97) under a geometry-free isotropic field, whereas $u_{\mathrm{var}}$ reverses the isotropic ordering to 0.76. Under the experimental prior, canonical $1/\nu_{\mathrm{can}}$ is a strictly increasing transform of $T=c/[\alpha(\gamma^2+2)]$ and is bounded above by $3+\sqrt{10}$. Thus ELVAE exposes a controllable sensitivity mechanism whose trained four-output realization is operationally informative, while the exact reconstruction-visible information remains three-dimensional and baseline image quality is a separate question.

Vision-Language-Action (VLA) models have shown strong capabilities in controlling robots across diverse manipulation tasks. However, their adversarial robustness remains largely underexplored, and exploiting this weakness can lead to physical-world harm. Existing attacks on VLA models often rely on pixel-space perturbations or white-box access, resulting in noticeable artifacts and limited deployability in real-world robotic systems. In this work, we propose DURA, a diffusion-based unrestricted robotic attack that generates visually natural adversarial patches for VLA models. DURA supports both white-box and black-box attack settings, where the black-box setting requires only the predicted actions of the victim model. By optimizing along the latent trajectory of a pretrained diffusion model, DURA generates visually natural patches while steering the robot toward attacker-specified target actions. Extensive experiments in both simulation and the real physical world show that DURA consistently outperforms existing methods. Our findings expose a safety risk for physically deployed VLA models and call for stronger defenses.

Large language models (LLMs) are increasingly used as relevance assessors in information retrieval (IR) evaluation, raising questions about how assessor framing affects judgment reliability and downstream system comparison. We study persona conditioning as a diagnostic mechanism for exposing LLM assessor sensitivity. Using task-oriented personas drawn from two complementary sources (PersonaHub and NVIDIA Nemotron-Personas-USA), we instantiate five assessor roles emphasizing intent interpretation, domain expertise, contrastive judgment, evidence verification, and global search-quality assessment, compared with a standard UMBRELA baseline. Across six LLM backbones on TREC DL20 and RAG24, our analyses reveal structured rather than uniform assessor sensitivity. Judgments usually remain close to the baseline while shifting assessment strictness, evidential threshold, or interpretation emphasis rather than producing widespread relevance reversals. At the system level, high-capacity models preserve system-ranking agreement, while smaller models amplify persona-induced instability. Local rank-displacement analysis shows sensitivity concentrates on particular retrieval systems and system types, especially neural ranking/reranking systems on DL20 and RAG-oriented pipelines on RAG24. Persona source matters less than assessor role and model capacity. These findings position persona-conditioned judging as a controlled sensitivity probe for stress-testing LLM-based IR evaluation pipelines and identifying systems whose evaluation outcomes are sensitive to assessor framing.

Universal approximation in reservoir computing is typically associated with a class of reservoirs. We show that universality can be associated with a single reservoir, considering a minimal setup of a single atom in front of a mirror. In its linear-transducer limit, our reservoir is a universal approximator of fading-memory maps under an operating class of checkable conditions, with a rate constant measured at the operating point. A given reservoir can reach arbitrary accuracy by changing measurement settings. The proof gives an explicit recipe: for a target accuracy, it specifies the required physical resources and resonator modes. Enlarging the number of accessible modes increases the matchable kernel span without reducing capability. Beyond the linear limit, the atom's saturation replaces high-order polynomial readouts, and the device operates on real-world tasks alongside classical baselines. Our results highlight an example of universality with a minimal quantum setup.

Volatility control converts risk estimates into portfolio exposure, yet existing approaches often rely on a fixed volatility estimator or a pre-defined control rule that may not adapt to changing market conditions. We propose VolRouter, a modular framework that formulates volatility control as state-conditioned routing over estimator-controller pairs. VolRouter first summarizes market conditions into a control-relevant state profile and then performs routing through three stages: state inference, switch review, and pair selection. The Router can be implemented using rule-based, learnable, or LLM-based decision modules, while portfolio actions remain generated by predefined control policies. We evaluate VolRouter across S&P 500, Multi-Asset, Bitcoin, and USDT volatility-control settings. VolRouter achieves the highest Sharpe ratio in three of four settings. On S&P 500, it improves Sharpe from 0.952 for RV + Naive Scaling to 1.222 while reducing maximum drawdown from 15.10% to 12.58% and daily CVaR from 1.76% to 1.32%. On Multi-Asset, it improves Sharpe from 1.498 to 1.540 and reduces CVaR from 1.56% to 1.18%. Bitcoin shows similar improvements in risk-adjusted performance, while USDT provides a boundary case where simpler state-aware selectors remain competitive. Ablation and sensitivity analyses show that the improvement comes from relative policy evaluation and selective persistent switching rather than simply expanding the policy library. These results suggest that volatility control can be viewed as a policy-selection problem when risk management requirements vary across market states.

Real-world data science involves long-horizon workflows that span data wrangling, exploration, modeling, visualization, and validation, and require coordinated use of tools such as notebooks, IDEs, terminals, browsers, and databases within real operating environments. Yet existing benchmarks lack real-computer interaction and do not evaluate whether agents can execute complete end-to-end data-science workflows in realistic computing environments, failing to capture the multi-stage, multi-tool nature of data-science practice. We introduce DSAgentBench, the first benchmark to evaluate whether agents can automate full data-science workflows inside real computer environments. DSAgentBench contains 275 diverse tasks covering the entire data-science life-cycle, reflecting the complexity and tool coordination required in practice. Each task requires grounding decisions in intermediate outputs and coordinated tool use, and includes a deterministic evaluator that verifies analytical correctness, visual outputs, and model performance rather than code-only execution. Our extensive experiments with 15 closed- and open-source models show that even the strongest agent, Claude-4.6-Sonnet, achieves only 56.70% task success, while all open-source agents remain below 1%, frequently failing at tool orchestration, OS grounding, and multi-step reasoning. These results reveal a substantial capability gap between current agentic systems and real data-science workflows, positioning DSAgentBench as a foundation for developing grounded, verifiable, autonomous data-science agents. We release DSAgentBench at https://github.com/vis-nlp/DSAgentBench.

AI agents can accelerate nutrition research, but their analyses inherit the identity, semantic, and release ambiguities of the underlying data. We present Nutrition Data Service (NDS), source-preserving infrastructure that operationalizes FAIR for automated use: description resolution makes release-specific records findable; typed crosswalks connect independently released resources; machine-readable interfaces expose versioned sources and crosswalks, supporting replayable and auditable analyses. On food-description benchmarks, NDS outperforms the best published language-model result on NutriBench. External and blinded crosswalk evaluations show that its typed contract favors defensible links and rejects unsupported mappings. In a person-level glycemic-index analysis, pinned NDS inputs produce identical outputs across models and repeated runs, while open-web reconstruction remains unstable. Together, these results show that agent-mediated nutrition research requires a new infrastructure that makes data identity, search, and crosswalk policy explicit.

Speculative decoding accelerates autoregressive large language model (LLM) inference by using a lightweight draft model to speculate multiple tokens, reducing expensive target model decoding steps. Its effectiveness depends heavily on draft selection, motivating adaptive methods that exploit variation across inputs and generation stages. On memory-constrained edge devices, however, these methods often fail to improve end-to-end throughput due to the overhead of switching between draft models. We identify a key limitation in this setting: the mismatch between draft selection and draft availability under tight memory budgets. To address this challenge, we present MemSpec, a prediction-guided, memory-aware runtime for adaptive speculative decoding on edge devices. MemSpec decouples draft selection from execution through proactive resident working-set management. A lightweight predictor estimates draft effectiveness from prompt and generation context, while a memory-aware scheduler reduces reactive model loading overhead. Experiments on a Jetson Orin Nano show that MemSpec improves steady-state generation throughput by 40.7% on average over state-of-the-art bandit-based adaptive methods while closely approaching the oracle upper bound.

Arabic maqam music microtonal, modal, and built on ornamented call and response is among the traditions most underserved by generative music models, whose training frameworks remain predominantly Western and equaltempered. Real time accompaniment sharpens this gap: an AI partner must listen, adapt dynamically, and respect idiomatic microtonal structures. Streaming text to music models provide strong generative capabilities but lack precise control interfaces. We present MazzikaAI, a knowledge based system that uses natural language as the actuator of a realtime control loop. By compiling live MIDI, gesture, and inferred harmony into continuously updated text prompts, MazzikaAI steers an unmodified streaming generator, Google Lyria RealTime, without requiring model finetuning. The system embeds expert knowledge of six core maqamat, characteristic ornaments, and ensemble dynamics, maintaining realtime responsiveness with subsecond keytoaudibleupdate latency. Empirical evaluations demonstrate that dynamic prompt compilation reliably grounds generation in microtonal scales, significantly increasing offgrid quartertone content over baseline generation. Beyond its core implementation, MazzikaAI illustrates how deterministic knowledgebased rules can effectively bridge expert, nonWestern musical traditions and unfinetuned foundation models. This architecture establishes a scalable paradigm for realtime humanAI cocreation, offering a generalizable blueprint for interactive accompaniment, adaptive music education, and culturally inclusive generative audio across diverse global idioms.

Long-horizon tool-using agents must reason over user goals, domain policies, tool calls, simulator state, and delayed verifiable rewards. Reinforcement learning (RL) is a natural fit for this setting, but multi-turn on-policy rollouts create long contexts, while model-specific attention layers may require custom masks and learned sink normalization. We present SINKFLEX-RL, a modular training system for RL in dual-control tool-use environments. The system combines a Gymnasium-compatible environment wrapper, a VERL-style rollout dataflow, group-relative policy optimization without a separate value model, and a sink-aware FlexAttention path designed to preserve model-specific sink scaling under causal and sliding-window masks. In a preliminary Tau2Bench retail run, validation reward (mean@1) rises from 0.25 early in training to $0.44$ later in the observed training window, while training-score and trajectory-reward proxies also trend upward. In a fixed-configuration memory benchmark, the optimized attention path reduces peak VRAM from 28.06GB to 22.52GB at 4096 tokens, a $19.7\%$ reduction, and runs the measured 8192-token configuration using $25.53$~GB where the eager baseline runs out of memory. These results illustrate the value of integrating environment interfaces, RL dataflow, and attention-kernel design for memory-feasible long-horizon agent training.

Although advancements in face landmark detection (FLD) methods continue to push performance boundaries, they overlook two major functional limitations: (1) different network parameters need to be trained independently for each “$N$-point” benchmark dataset, and (2) a model trained on an “$N$-point” dataset reliably outputs only the $N$ landmarks. In our work, we first conceptualize Face Part-Anchored Landmark Positions (FPALPs), wherein each landmark is treated as a progression value between zero (start) and one (end) along a face part's contour. Every landmark can be expressed in the FPALP format, irrespective of its source dataset, hence unlocking the ability to unify all “$N$-point” datasets into a single dataset. Secondly, we represent each landmark with an FPALP-based query, refine it progressively with a cross-modality decoder, and predict its coordinates based on the final representation. Our approach, called Unified Dynamic FLD, embodies these two design choices and streamlines the landmark detection pipeline by enabling (1) a single model to learn on any number of “$N$-point” datasets, and (2) yield any number of specific landmark predictions by loading the designated landmark queries at runtime. Extensive experiments on multiple benchmark datasets show that our method delivers these benefits while remaining competitive with, and in several cases outperforming existing state-of-the-art methods.

Hospital quality improvement (QI) programs routinely face multiple candidate interventions to optimize hospital flow, but existing methods struggle to estimate and rank the causal effects of such interventions. This work focuses on one of the most standard hospital metrics, the average length of stay (LOS), and its causal estimand, the average time saved. To characterize this causal effect, qualitative approaches rely on expert judgment to map patient trajectories, making them susceptible to cognitive biases; quantitative approaches rely on data-driven models, which fail when interventions are hypothetical with no historical data or have complex causal mechanisms that require clinical reasoning rather than data alone. We propose expert-guided g-computation, or egg-computation, which combines the complementary strengths of both approaches by connecting the Gantt charts commonly used to map patient trajectories with the causal DAG literature. We introduce a causal model over Gantt charts and establish identification using a variant of g-computation that seeks expert input only for components unidentifiable from data. To make egg-computation practical, we develop an LLM-assisted pipeline that reliably scales up expert reasoning. In simulations, egg-computation outperforms conventional causal inference methods when patients have diverse causal structures and intervention mechanisms. In a study of eleven candidate QI interventions at an urban safety-net hospital, the LLM pipeline generated graphs and time-saving estimates highly concordant with those of human experts. Beyond healthcare, egg-computation is a broadly applicable framework for estimating the average time saved for candidate interventions whose causal mechanisms can be represented using Gantt charts.

We introduce narrative keyframing, an interaction technique for AI-assisted creative writing that lets writers specify different types of narrative constraints at selected moments in a story, then use AI to generate intervening prose. Inspired by the use of keyframing in animation, narrative keyframing offers a flexible way to connect story planning with adaptive control over generated text. We explore three types of keyframes: plot keyframes define significant events in a story, character keyframes represent how individual characters change over the narrative, and perspective keyframes capture how individual characters experience different events through first-person narratives. Plot and character keyframes offer a flexible way to adapt the type of high-level conditioning explored in previous AI writing tools to more customizable, iterative, and fine-scale control, while perspective keyframes add a new way to control characterization and focalization by using first-person narratives as an intermediary. Through a user study, we show that narrative keyframing supports a more controllable, transparent, and engaging way to use generative AI in creative writing.

AI agents are increasingly being developed to assist humans in various applications, and Large Language Models and other deep network architectures are considered to be state of the art for such agents. These methods are impressive stochastic predictors, but they are resource-hungry, opaque, and known to make arbitrary decisions in novel situations due to the narrow set of underlying representation and processing choices. Our work seeks to explore the design of architectures for such AI agents based on core principles that can be traced back to the early pioneers of AI but are not fully utilized in modern AI methods. We do so in this paper in the context of the core problem of AI agents addressing ambiguity in the objects being referred to by the human participants. Humans address such ambiguity by heuristically leveraging compositional knowledge of domain context and the preferences of the other human participants. Drawing inspiration from this observation, we describe an architecture that embeds the principle of hierarchical compositionality and uses simple heuristics to achieve the desired disambiguation. Specifically, domain objects are represented in terms of primitive attributes drawn from human-validated semantic feature norms, and a hierarchical combination of attributes and concepts automatically identified from a limited observed history of interactions of an assistive agent with specific users. The assistive agent then achieves the desired disambiguation by reasoning with knowledge of this compositional hierarchy; axioms governing domain dynamics; and models of semantic compatibility, session salience, and user-specific thematic preference, requesting human clarification when necessary. Experiments show that our approach consistently outperforms state of the art data-driven baselines, supporting adaptation to specific user profiles.