arXiv subjects

Yitong Li

Publications and source records attributed to Yitong Li.

At least 19 recordsLinked to original sources

arXiv2026

MoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health Coaching

Despite rapid advances in large language models (LLMs), their deployment in the sports and health domain remains high-risk due to the absence of domain-specific evaluation benchmarks. Existing general-purpose benchmarks (e.g., MMLU and HELM) primarily assess broad knowledge and reasoning capabilities, but fail to capture critical domain requirements such as exercise safety, physiological appropriateness, personalization, and actionable coaching quality. This gap is particularly concerning in fitness and health applications, where incorrect or unsafe recommendations can lead to physical harm. To address this limitation, we introduce MoveBench, the first comprehensive benchmarking framework specifically designed for evaluating LLMs in sports and health coaching. The proposed benchmark integrates heterogeneous task formats, including multiple-choice questions (MCQs) and open-ended question answering (OpenQA), grounded in real-world user scenarios and paired with structured rubrics emphasizing safety compliance, scientific validity, and training logic. Through large-scale evaluation of 15 diverse LLMs on 809 questions with 5,012 rubric-level criteria, our results reveal substantial performance variability, particularly in safety awareness, structured program design, and personalization capability, highlighting a critical gap between knowing and doing. Our work establishes a foundational evaluation standard for sports and health LLM applications and provides a reproducible framework for advancing safe, reliable, and context-aware AI-driven coaching systems. Open resources are available at https://github.com/SatoSakula/MoveBench.

cs.AI↗

arXiv2026

Scaling Online Exploration for Robot Agents

Robot learning has advanced by scaling demonstration datasets and model capacity; however, deploying a new robot embodiment or task typically necessitates a dedicated teleoperation corpus. The resulting policies remain bounded by their imitation distributions, limiting generalization and continual self-improvement. We introduce RAX (Robot Agent eXploration), an open-source framework for robot agents centered on scaling exploration: enabling agents to improve by acquiring online experience and synthesizing executable programs. In RAX, a general-purpose coding agent generates robot control code, operating without action demonstration data, hand-engineered action primitives, or value-function guidance. Through online exploration, the agent identifies decision sequences that achieve task goals and compiles them into modular programs encapsulating the full perception-planning-control loop. Using RAX, we present a systematic multi-embodiment study on scaling online exploration across precision and long-horizon manipulation tasks using single-arm, bimanual, dexterous, mobile, and humanoid systems in simulation and on real robots. Generalization performance on unseen environments improves consistently as exploration scales. In simulation, RAX outperforms top baseline models by +37.4% on average. On real robots, RAX autonomously learns memory-dependent, articulated, high-precision, deformable, and long-horizon mobile manipulation tasks without human demonstrations, continually improving performance by composing verified skills from prior exploration sessions. As exploration scales, we observe emergent capabilities including active perception, error recovery, safety guardrails, and human-robot collaboration. We open-source RAX as a generalist robot-agent platform in which new robot embodiments are integrated through their APIs alone, without teleoperated data collection or policy training.

cs.RO↗

arXiv2026

SoL-Refiner: Speed-of-Light One-Step Refinement for High-Resolution Video

High-resolution video generation is expensive, as its cost grows rapidly with the number of spatiotemporal tokens. A practical alternative first generates a lower-resolution video and then applies a refiner, but conventional multi-step refinement introduces a second sampling bottleneck. We present SoL-Refiner, a one-step video refiner that transforms low-resolution model outputs into 4K videos with a single denoising step. Our three-stage recipe combines high-resolution continual training, reinforcement learning (RL) post-training, and a final one-step distillation. We introduce Refiner-Bench, a video refinement benchmark constructed from the outputs of different video generators, and use a shared-input protocol to compare refiners at approximately 2K output resolution. At 2K, the one-step SoL-Refiner outperforms all external refiners on the VBench and UniPercept averages, while at $3840\!\times\!2176$ it improves both metrics over the three-step LTX-2.3 Refiner. With the complete acceleration stack, SoL-Refiner achieves an $8.91\times$ speedup in refinement latency over the same baseline in our 2K latency setting.

cs.CV↗

arXiv2026

Sol-H3: Recursive Self-Improvement for MiniMax-H3 Inference Acceleration on Sol-Engine across Cloud and Edge

Video diffusion models are rapidly scaling and exhibiting enhanced generation capabilities. Among these recent advancements, MiniMax-H3 stands out as a highly capable, production-level open-source model. However, its 33-billion parameters and multi-step iterative denoising process introduce substantial computational overhead. Consequently, their practical production is hindered by generation latency in the cloud deployment like NVIDIA-GB200, alongside strict memory limits that pose further challenges at the edge device like DGX-Spark. To address these diverse hardware bottlenecks from cloud to edge device, we present a full-stack inference pipeline that integrates efficient algorithmic design with optimized operator implementations. Algorithmically, we introduce a cross-resolution two-stage generation scheduler that exploits the step-wise nature of diffusion: early low-resolution steps rapidly establish the global layout, while later high-resolution steps focus refinements of local and perceptual details. These stages are connected by a learned latent-to-latent mapping module, completely eliminating the computationally expensive VAE decode-reencode cycle for resolution transferring cross different resolutions. For operator implementation, we deploy a Recursive Self-Improvement (RSI) loop that searches kernel fusions and memory layouts, evaluating latency together with numerical agreement. Together, these optimizations deliver up to 30x end-to-end speedup and 20% lower memory: a 5-second 1344x768 video with audio is generated 3.5x faster than real time on an 8xGB200 node, and in under a minute fully memory-resident on a single DGX Spark.

cs.AI↗

arXiv2026

DraftAttention2: Fast Video Diffusion with Low-Resolution-Guided Mixed-Precision Attention

Video generation has broad applications in content creation and entertainment. Diffusion transformers have advanced the quality of generated videos, but attention over spatiotemporal tokens becomes increasingly expensive as video resolution and duration increase. We present DraftAttention2, a training-free framework that uses the low-resolution draft attention map to jointly select attention blocks and assign their numerical precision. Specifically, spatial 2D average- and max-pooled queries and keys capture complementary regional statistics to estimate block importance, and a shared ranking assigns higher precision to important blocks, lower precision to less important retained blocks, and skips the rest under configurable budgets. Our analysis separates sparsification error from attention-weighted quantization error, establishing when recovering skipped interactions with low-bit computation tightens the output-error bound. This analysis motivates retaining more interactions at low precision while reserving higher precision for blocks with larger attention mass. To translate these fine-grained assignments into practical speedups, we further develop fused operand preparation and a single attention kernel with precision-specific phases, sharing data movement, softmax statistics, and output accumulation across precisions. Experiments demonstrate that our method achieves a superior quality-efficiency trade-off over existing efficient video generation methods. Notably, its advantage is particularly pronounced for few-step video diffusion, where jointly combining sparsity with 4- and 8-bit mixed-precision computation substantially improves generation quality while retaining significant acceleration. Code is available at https://github.com/anemoi-project/anemoi

cs.CV↗

arXiv2026

Electrically reconfigurable dipolar polaritons with highly tunable nonlinearity in a homobilayer microcavity

Active control over optical nonlinearity in solid-state systems is central to unlocking exotic many-body phenomena and photonic devices. While exciton-polaritons in transition metal dichalcogenides (TMDs) offer a promising platform, their practical utility is impeded by fixed interactions and a trade-off between nonlinearity and oscillator strength. Here, we report electrically tunable dipolar polaritons in a dual-gated bilayer MoS2 microcavity, demonstrating in situ reshaping of the dispersion and modulation of the light-matter coupling strength via the quantum-confined Stark effect. Crucially, this electrical control yields a seven-fold enhancement of the polariton-polariton interaction strength. This enhancement arises from the combined tuning of the effective exciton-exciton interaction and the excitonic Hopfield coefficient. In addition, electrostatic doping provides an independent knob to continuously drive a strong-to-weak coupling crossover. Our findings establish dual-gated TMD homobilayer as a versatile platform for on-chip, dynamically reconfigurable nonlinear light-matter physics.

physics.optics↗

arXiv2026

A Reconfigurable Dual-Opposition Architecture for Single-Hand Assembly and Manipulation

In-hand assembly is constrained by the need to maintain grasps on two separate parts while controlling their relative motion within a single hand. To enable both in-hand assembly and manipulation, we present a reconfigurable dual-opposition architecture. Specifically, to support simultaneous grasping of two parts and coordinated in-hand manipulation, four independently actuated fingers are organized into two virtual finger (VF) oppositions, with their relative configuration controlled by a reconfigurable palm. To describe hand motion and simultaneous two-object grasping configurations, a kinematic model of the fingers and palm and an object-size-conditioned workspace formulation are built. To further evaluate motion performance and assembly capability, finger-joint motion and palm tracking are characterized, and in-hand assembly is demonstrated through tasks involving grasping, alignment, fastening, and pressing. Ablation experiments further demonstrate the importance of finger abduction/adduction and palm reconfiguration for successful in-hand assembly. In simulation, the proposed hand achieves a mean continuous sphere rotation success rate of 98.6% over diameters of 40-230 mm, compared with 73.8% for the LEAP Hand. After policy fine-tuning with external disturbances, the proposed hand achieves 92.8% success under disturbances from multiple directions, compared with 45.2% for the LEAP Hand. Hardware demonstrations further show in-hand rotation of objects of different sizes using policies trained in simulation. Together, these results show that the proposed architecture supports both assembly of two separately held parts and coordinated manipulation of a single object within one hand.

cs.RO↗

arXiv2026

Bridging Modalities on the Cortex: Surface-based MRI to PET Translation with a Diffusion Bridge

Cortical hypometabolism measured by Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) is a highly sensitive biomarker for dementia diagnosis. However, high costs, radiation exposure, and limited accessibility constrain its clinical utility. While cross-modal synthesis from Magnetic Resonance Imaging (MRI) offers a promising alternative, existing volumetric generation methods do not explicitly account for the highly folded cortical geometry, where disease-related patterns predominantly reside. To address this, we introduce a novel surface-based diffusion bridge framework DB-SUiT for MRI-to-PET translation that operates natively on the cortical manifold. A conditional Spherical U-shaped vision Transformer (SUiT) is specifically designed to model the intricate cross-modal relationships while preserving surface topology. It combines spherical convolutional encoders for multi-scale surface feature extraction with bottleneck Transformers to capture long-range spatial dependencies, while incorporating demographic and subcortical conditions to refine the synthesis. Evaluated on two datasets, including subjects with different dementia types, DB-SUiT demonstrates high-fidelity synthesis that substantially outperforms other baselines. In automated dementia classification, synthesized PET surfaces improve performance over MRI by 14.2% and PET volumes by 11.3%, approaching the performance of real PET surfaces. In a blinded reader study, synthetic PET achieved 85.5% diagnostic accuracy, compared with 75.8% for MRI and 95.2% for real PET. This further demonstrates cross-cohort and cross-pathology generalization, as the model was evaluated without retraining on an external cohort that included a dementia subtype not represented during training. Our code is available at https://github.com/ai-med/DB-SUiT.

cs.CV↗

arXiv2026

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness

As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement. We take an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments for harness rollouts. At this scale, the process yields reusable improvements that transfer beyond their development setting, moving automated harness discovery toward production-level outcomes. Four mechanisms survive selection and form SoL-Pi, spanning action execution, context compaction, observation handling, and delegated reading. On the 51-task EdgeBench evaluation, SoL-Pi achieves performance comparable to Pi across GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7-49.0% and API cost by about one third. In other words, estimated hourly savings are \$8.75-\$13.50 relative to native Codex and Claude Code harnesses, and \$4.36-\$5.71 relative to Pi.

cs.AI↗

arXiv2026

Generating Reports or Repeating Templates? Measuring and Mitigating Template Collapse in 3D CT Report Generation

Modern 3D medical vision-language models (VLMs) can generate fluent radiology-style text while exhibit critically low pathology detection and output diversity, collapsing to generic templates that under-report rare yet critical findings. We identify this failure mode as Template Collapse. This failure stems from the unique constraints of 3D medical imaging, e.g., limited data, severe label imbalance, and weak signals from volumetric encoders. Under these constraints, text-generation objectives encourage shortcut learning and fluent but weakly grounded reports. We systematically diagnose the Template Collapse through clinical fidelity, output diversity, normal-template bias, and rare-finding survival. To mitigate it, we propose CLarGen, a decoupled framework that separates what to say (clinical detection) from how to say it (language synthesis). CLarGen uses (i) a Latent Query Transformer for multi-label pathology detection, (ii) pathology-guided retrieval for clinically matched exemplars, and (iii) a medical language model to synthesize the final report from detected findings and retrieved context. Across state-of-the-art 3D CT report generation baselines, CLarGen mitigates Template Collapse and substantially improves clinical accuracy (macro-F1 0.487 vs. 0.189; CRG 0.472 vs. 0.368) while maintaining fluent reporting. Our results suggest that explicit, measurable clinical grounding is essential for template-collapse-resistant 3D CT report generation. Code is available at https://github.com/ai-med/CLarGen.

cs.CV↗

arXiv2026

MARS: Multipath Adaptive Reliable Service

Multipath transport is important for Internet/WAN services that move data volumes across heterogeneous paths, including geo-distributed analytics, content distribution, and cloud-service pipelines. Existing solutions face a trade-off: end-to-end transports such as MPTCP and MPQUIC are deployable but limited by endpoint-visible paths and delayed congestion feedback, while routing- or forwarder-assisted approaches often require infrastructure support or lack safe coordination across forwarding choices. This paper presents MARS, a receiver-driven, forwarder-assisted multipath transport. MARS combines tier-synchronized overlay path discovery with coupled consumer/forwarder congestion control, enabling it to expand usable forwarding opportunities and react near bottlenecks. It runs as an incrementally deployable UDP overlay at clients, servers, relays, or CDN-like nodes. We implement MARS in simulation and as a prototype, and evaluate it through simulation and Mininet emulation across deployment scopes, loss rates, and a forwarding-face outage scenario. Results show MARS provides deployment-dependent benefits: with endpoint-only deployment, it performs comparably to the evaluated ECMP-limited configurations of MPTCP and MPQUIC. With cooperating overlay forwarders, it expands the usable path set from routing-exposed forwarding candidates. Across the tested loss conditions, it reduces maximum T95 by up to 66.7% and 63.9% relative to the evaluated path-expanded MPTCP and MPQUIC configurations, respectively, given the same path set. Path discovery remains lightweight, flow fairness remains high, and MARS degrades gracefully during an emulated forwarding-face outage and recovers quickly after face restoration. Overall, ICN-style receiver-driven forwarding can serve as a deployable overlay transport substrate for coordinated WAN multipath without requiring changes to IP routing.

cs.NI↗

arXiv2026

Conditional Diffusion for 3D CT Volume Reconstruction from 2D X-rays

Computed tomography (CT) provides rich 3D anatomical detail but is often constrained by high radiation exposure, substantial costs, and limited availability. Standard chest X-rays are cost-effective and widely accessible, but provide only 2D projections with limited pathological information. Reconstructing 3D CT volumes from 2D X-rays could markedly increase diagnostic accessibility, yet existing methods rely predominantly on synthetic X-ray projections, limiting clinical generalization. We propose AXON, a multi-stage diffusion-based framework that reconstructs 3D CT volumes directly from real X-rays with substantially improved fidelity over existing approaches. AXON follows a coarse-to-fine paradigm: a Brownian Bridge diffusion model first captures global anatomical structure, and a ControlNet-guided refinement stage then enhances local intensity detail and anatomical realism. To alleviate the depth ambiguity inherent in 2D-to-3D reconstruction, AXON incorporates bi-planar X-ray views, enabling more accurate spatial reasoning and structural recovery. A dedicated super-resolution module further increases the spatial resolution of the generated volumes. Experiments on public and external datasets show that AXON consistently surpasses state-of-the-art approaches while generalizing across diverse clinical distributions. At our highest-resolution bi-planar setting, AXON achieves an 11.9% improvement in PSNR and an 11.0% increase in SSIM over the strongest baseline evaluated at that resolution. In the $128^3$ single-planar setting, it maintains a lead of 7.8% in PSNR on LIDC-IDRI, with larger margins on the external clinical dataset of 8.0% in PSNR and 16.9% in SSIM. Our code is available at https://github.com/ai-med/AXON/.

cs.CV↗

arXiv2026

Video2DoorTraversal: Push Door Traversal via Simulated Door Twins

Door opening and traversal is a long-horizon loco-manipulation task that requires precise handle interaction and coordinated base-arm control. We present Video2DoorTraversal, a single-video real-to-sim-to-real framework for wheel-legged mobile manipulators. Given one RGB video of a real door, DoorTwin reconstructs an instance-aligned, articulated, and simulation-ready door twin with realistic geometry and appearance. A simulation-in-the-loop agent converts the recovered articulation into a parameterized skill program and iteratively refines failed rollouts to generate physically executable demonstrations. These demonstrations are used to train ArticuACT, a dual-depth policy that predicts coordinated base, arm, and gripper commands using robot-centric camera conditioning and interaction-aware supervision. With all perception and policy inference running onboard, the system achieves a 96.57% average success rate across five real doors and an 80.95% zero-shot success rate on structurally similar unseen doors, while completing the full approach, opening, and traversal sequence in approximately 13s on average. Project Page: https://video2doortraversal.github.io/.

cs.RO↗

arXiv2026

MultiMoQ: Multi-Access Media-Over-QUIC for Robust Immersive Video Streaming

Live immersive video streaming, particularly 360-degree video, is increasingly adopted in applications such as virtual events, sports broadcasting, and remote education. Existing approaches struggle to support high-bitrate immersive streaming for large numbers of concurrent users, with coarse-grained delivery limiting responsiveness and insufficient support for coordinating concurrent tile streams. Media over QUIC (MoQ) has recently emerged as a promising solution for large-scale media delivery, yet it lacks robustness under bandwidth-constrained conditions, often resulting in playback stalls. To address these challenges, we present MultiMoQ, a multi-access tile streaming framework built on MoQ that redesigns its delivery mechanism for robust high-bitrate streaming across multiple access paths while supporting flexible tile scheduling and seamless access switching. We implement a fully functional prototype of MultiMoQ and evaluate it in network emulation under heterogeneous real-world network conditions, comparing against Dynamic Adaptive Streaming over HTTP (DASH) and standard MoQ. Results show that MultiMoQ increases goodput for enhancement tiles and base video and reduces enhancement-tile tail end-to-end latency relative to DASH, while preserving audio continuity and avoiding the persistent stalls of standard MoQ. These transport gains translate into smoother viewport playback, and the ablation results further confirm the contribution of multi-access control to playback continuity.

cs.NI↗

arXiv2026

JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion

Real-time video editing requires low-latency causal generation with bounded computational resources while preserving source fidelity and long-term temporal consistency. We present JoyAI-Video-Edit, a 16B-parameter autoregressive diffusion framework for real-time, open-ended video editing without access to future frames or a predefined video duration. Our method combines chunk-wise autoregressive adaptation, Source-Anchored Distribution Matching Distillation (SA-DMD), and Long-Horizon Autoregressive Distillation to reduce train--inference mismatch, preserve source fidelity during two-step generation, and mitigate accumulated temporal drift. Extensive automatic and human evaluations show that JoyAI-Video-Edit substantially outperforms existing streaming editors and remains competitive with strong offline systems on both short and long videos. The complete system achieves end-to-end 720p video editing at approximately 30 FPS on a single Nvidia B200 GPU. Code is available at https://github.com/jd-opensource/JoyAI-Video-Edit.

cs.CV↗

arXiv2026

Piggybacking on Perception: Stealthy Concurrent Audio Prompt Injections against Multimodal LLM Agents

Large Language Model (LLM)-driven multimodal agents are increasingly deployed to execute autonomous tasks via continuous audio interaction. While this paradigm enhances interaction naturalness, it introduces a critical yet under-explored attack surface, as audio inputs inevitably contain environmental noise beyond user control. In this paper, we investigate concurrent audio prompt injection attacks targeting multimodal agents. Distinct from traditional acoustic attacks on voice devices, we propose novel techniques for instruction augmentation and scenario concealment. These methods allow malicious audio instructions to imperceptibly "piggyback" onto user speech, thereby hijacking agents to execute malicious actions. To systematically quantify this threat, we construct AudioAgentSecurity, the first comprehensive benchmark for audio instruction injection attacks, encompassing 8 real-world task scenarios and 10 distinct attack patterns. We evaluate 11 state-of-the-art agents, including Gemini 3 Pro and GPT-4o-audio. Notably, our methods achieve an average Attack Success Rate (ASR) of 69.10\% against the advanced Gemini 3 Pro. To counter this threat, we further introduce Cascaded Audio Decoupling and Verification (CADV), a defense mechanism based on source separation and consistency analysis. Compared with existing prompt-level defenses, CADV achieving up to 96\% detection accuracy and providing effective protection against a broad range of acoustic injection attacks. Finally, real-world experiments with human volunteers on Doubao AI Smartphone in diverse dynamic real-world scenarios confirm the attacks' high stealth and efficacy, while demonstrating that our defense reliably mitigates these vulnerabilities.

cs.CR↗

arXiv2026

StairMaster: Learning to Conquer Risky Hollow Stairs for Agile Quadrupedal Robots

Climbing hollow stairs remains a challenging problem for quadruped robots due to the high risk of leg trapping, severe depth sparsity, and high-frequency depth-sensing noise. In this paper, we propose StairMaster, a novel three-stage reinforcement learning framework for stable locomotion on such extreme discontinuous terrains. Our architecture integrates a Cross-Attention mechanism to extract structural features from noisy depth data, alongside a Spatial-aware Recurrent Unit (SRU) that maintains robust spatio-temporal memory to mitigate perception blind spots. To bridge the sim-to-real gap in depth perception, we propose a high-fidelity sim-to-real depth sensor modeling pipeline that faithfully replicates real-world sensor artifacts. Additionally, we employ a 3D waypoint-guided active perception reward for proactive sensing, alongside hollow gap kinematic and stair edge penalties to ensure precise foothold placement. We successfully deployed StairMaster on a Unitree Go2 robot, demonstrating its ability to conquer hollow stairs with an unprecedented incline of up to 55$^\circ$ through zero-shot transfer. To the best of our knowledge, this is the first RL-based policy to achieve such steep hollow stair climbing in real-world environments. Project Website: https://sivan666666.github.io/StairMaster/.

cs.RO↗

arXiv2026

Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification

Diffusion transformers are essential for high-fidelity video generation, but long token sequences make attention a dominant inference bottleneck. Training-free dynamic sparse attention alleviates this bottleneck by computing only selected key-value blocks, yet existing methods struggle to sparsify attention both efficiently and accurately for two reasons: (1) Rigid, unpredictable, and costly routing: selecting a fixed fraction of top-ranked blocks by proxy score imposes fixed budgets, whereas retaining blocks to reach a target cumulative proxy probability mass yields dynamic but potentially imbalanced budgets; both incur non-negligible overhead from computing and materializing proxy scores. (2) Lossy keep-or-drop sparsification: unselected blocks are discarded entirely, degrading accuracy under aggressive sparsity. These limitations motivate cheaper dynamic-budget routing while limiting accuracy degradation. In this paper, we introduce training-free Sol-Attn (Sparsifying online attention), which unifies dynamic routing, sparse computation, and approximation correction in a single online-softmax pass, achieving a better accuracy-efficiency trade-off in sparse attention. The core of Sol-Attn is on-the-fly block thresholding with proxy-score reuse, which selects critical blocks by comparing block proxy scores against a threshold during online softmax. This design enables dynamic yet controllable block budgets without materializing the proxy map, while directly reusing the proxy scores of unselected blocks to approximate their contribution. Experiments across image and video generation tasks show that Sol-Attn advances the quality-efficiency frontier of training-free sparse attention, delivering 2.1 times and 2.3 times end-to-end speedups for video generation and editing, respectively, while preserving visual quality.

cs.CV↗