arXiv subjects

Yunke Wang

Publications and source records attributed to Yunke Wang.

At least 19 recordsLinked to original sources

arXiv2026

Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied Manipulation

Vision-language models can interpret instructions and plan manipulation tasks, yet struggle when acting directly as robot policies. On RoboDojo, a large planning-execution gap separates high-level task decisions from physical execution: removing arm execution raises GPT-6 Astra's success rate from 22.5% under the reference control interface to 76.2%. We present Auriga, a task-agnostic agent harness that closes 68.3% of this gap in zero-shot manipulation without target-task trials or resets. In Auriga, the model makes all task decisions over an explicit task belief and acts through either direct tool calls or synthesized multi-step Python programs, while the harness supplies calibrated 3D measurements, finger-referenced dual-arm control, and a single admission gateway to the robot. Before motion, the model inspects a visual trajectory preview while simulation is paused; during execution, the harness checks declared stage tolerances and grasp bounds, stopping the robot on divergence and returning the measured discrepancy for in-place recovery. Developed across 13 cross-task capability iterations and frozen for evaluation, Auriga ranks first on the RoboDojo leaderboard with a 57.7% success rate (64.0 Score), more than doubling the reference harness on the same model, and remains first under scene randomization while learned policies drop by up to 33.9%. On RoboCasa365, the same frozen harness reaches a 73.2% success rate, outperforming VLA-assisted agents by up to 23.1% on composite tasks. Project page: https://auriga.stelledge.com/ Code: https://github.com/lose4578/Auriga

cs.RO↗

arXiv2026

Prepare to Turn: Learning Anticipatory Maneuvering for Fast Sequential Quadrupedal Navigation

Fast quadrupedal navigation requires preparing for the next turn while approaching the current goal. Policies trained to reach and remain at a single target can instead produce stop-and-go motion when applied to sequential goals, compromising its agility potential. We introduce PrepTurn, a reinforcement learning framework for anticipatory maneuvering in sequential local navigation. PrepTurn maps the current and next pose goals directly to joint actions, using the upcoming goal to inform how the robot should approaches the current one. Its persistent-credit reward encourages timely completion of the whole sequence by continuing to reward reached goals while moving on and preparing subsequent maneuvers, and an automatic curriculum that gradually increases goal difficulty is critical for learning stable turning behaviors. Together, these components enable the policy to coordinate approach speed and reorientation across goal transitions without prescribed trajectories or arrival times. Simulation and onboard Unitree Go2 experiments demonstrate fast, efficient sequential traversal with high completion and reduced stop-and-go motion. We further show that when the goal-reaching tolerances are more relaxed, the policy learns to exploit the freedom allowed by the task with emergent behaviors, such as turning to the next goal in advance and cutting corners within the permitted regions behaviors, to traverse goal sequences more efficiently.

cs.RO↗

arXiv2026

OccamView: Object-Conditioned View Selection for Frame-Budgeted Active 3D Gaussian Reconstruction

Active 3D Gaussian reconstruction fundamentally relies on selecting informative next-best views under limited sensing budgets. Existing active 3DGS methods primarily plan viewpoints according to geometric information gain, treating object-induced hidden regions in the same manner as general unexplored space. Under tight frame budgets, such geometry-driven strategies may prioritize global scene coverage while leaving partially observed objects incompletely reconstructed. To address this limitation, we propose OccamView, an object-conditioned view-selection framework for frame-budgeted active 3D Gaussian reconstruction. Rather than predicting unseen object geometry or performing shape completion, OccamView maintains an online object memory from open-vocabulary detections grounded in measured RGB-D observations and represents unresolved local occupancy around detected objects as conservative hidden-region proxies. Candidate viewpoints are then evaluated using an occlusion-aware proxy-coverage score. Furthermore, we introduce a Geo-Floor mechanism that restricts object-conditioned re-ranking to geometrically competitive candidates, allowing object-conditioned cues to guide complementary observations while preserving the geometry-driven exploration behavior of the underlying planner. Experiments on Replica and Matterport3D under a unified frame-budgeted protocol show that OccamView consistently reduces Completion and improves Completion Ratio across five frame budgets, with particularly pronounced gains under limited frame budgets. These results demonstrate that lightweight object-conditioned cues effectively complement geometry-driven active view planning.

cs.RO↗

arXiv2026

Early Semantic Grounding in Image Editing Models for Zero-Shot Referring Image Segmentation

Instruction-based image editing (IIE) models have recently demonstrated strong capability in modifying specific image regions according to natural language instructions, which implicitly requires identifying where an edit should be applied. This indicates that such models inherently perform language-conditioned visual semantic grounding. In this work, we investigate whether this implicit grounding can be leveraged for zero-shot referring image segmentation (RIS), a task that requires pixel-level localization of objects described by natural language expressions. Through systematic analysis, we reveal that strong foreground-background separability is already present in the internal representations of these models at the earliest denoising timestep, well before any visible image transformation occurs. Building on this insight, we propose a training-free framework that repurposes pretrained image editing models for RIS by exploiting their intermediate representations. Our approach decomposes localization into two complementary components: attention-based spatial priors that estimate where to focus, and feature-based semantic discrimination that determines what to segment. By leveraging feature-space separability, the framework produces accurate segmentation masks using only a single denoising step, without requiring full image synthesis. Extensive experiments on RefCOCO, RefCOCO+, and RefCOCOg demonstrate that our method achieves superior performance over existing zero-shot baselines. Project page: \url{https://jessie459.github.io/EarlySemanticGrounding/}.

cs.CV↗

arXiv2026

Latent World Models Should Plan with Distances Along the Latent Manifold

Latent world models such as LeWM plan by searching for an action sequence whose predicted terminal latent is closest to the goal latent under an ambient norm, such as the Euclidean distance. Planning with such a cost assumes that proximity in the ambient latent space reflects proximity to the goal in the environment. Studying LeWM, we find that this assumption holds only in a small neighborhood around the goal, whereas farther away the L2 cost forms spurious low-cost basins that trap the planner. Yet neighborhood-based embeddings show that the latents retain the underlying physical state. We therefore hypothesize that the latents lie near a curved, low-dimensional manifold, and that the terminal cost should measure the distance along this latent manifold rather than through the ambient space. To estimate this manifold distance, we propose the Neighborhood Graph Cost (NGC), which computes shortest-path distances on a mutual nearest-neighbor graph constructed over training latents. NGC changes only the terminal cost and requires no retraining. This single change improves the planning success rate of LeWM across all four benchmark tasks and nearly solves TwoRoom and Reacher: even at the farthest goal offsets tested, NGC reaches 95% and 100% success, whereas the L2 cost falls to 27% and 81%. These results support the manifold hypothesis and our argument that latent world models should plan with distances along the latent manifold.

cs.LG↗

arXiv2026

VDOT++: Unified Few-Step Video Generation via Unbalanced Optimal Transport Distillation

Video creation spans text-to-video (T2V), image-to-video (I2V), and condition-based generation, yet video diffusion models remain costly because they repeatedly evaluate large backbones during sampling. Distribution matching distillation (DMD) reduces this cost, but its reverse Kullback--Leibler (KL) objective can provide unstable or incomplete guidance when the student and teacher distributions have limited overlap. VDOT addressed this issue by adding optimal transport distillation (OTD), whose explicit coupling supplies geometric directions for condition-based generation. Balanced OTD, however, performs full-mass matching between the spatial tokens of each corresponding student--teacher frame pair. This assumption weakens for T2V and I2V, where one condition admits many valid outputs and spatial content need not align across different realizations. We present VDOT++, a unified distillation framework that applies the same training recipe separately to generators for the three task families. It makes OTD robust to output diversity through an asymmetric unbalanced formulation that allows unreliable student tokens to carry less mass while maintaining coverage of the teacher tokens. An $\ell_1$ ground cost further replaces mean-based aggregation with a more mode-preserving weighted median that limits the influence of distant transport targets. The two changes respectively determine whom to match and how the selected targets should be aggregated. We additionally combine distribution matching and adversarial refinement through sequential backward passes, and exploit the decoupled score networks for cross-scale distillation, where larger score networks improve a compact generator. Experiments on UVCBench, VBench, VBench-I2V, and the VACE benchmark show that the resulting four-step generators are competitive with many-step teachers and strong few-step baselines across all three task families.

cs.CV↗

arXiv2026

ShelfChange3D: Object-Level 3D Change Detection for Retail Shelf Monitoring

Reliable shelf monitoring is an important capability for retail automation, yet existing out-of-stock detection methods mainly operate in image space and lack metric 3D localization for downstream robotic systems. We formulate shelf monitoring as object-level 3D change detection: given two RGB-D observations captured at different times, the goal is to identify changed products and localize each change with a 3D bounding box. To support this task, we introduce ShelfChange3D, comprising 145K synthetic and 5K real-world paired RGB-D observations with object-level 3D change annotations. We further propose ChangeBox, an end-to-end framework that jointly reasons over paired observations and predicts object-level 3D change boxes. To improve localization accuracy, we introduce a geometry-based refinement stage that exploits depth and gravity prior to estimate relative pose and refine predicted boxes. Experiments show that ChangeBox outperforms existing change detection baselines, with further gains from refinement and effective transfer from synthetic to real-world observations.

cs.CV↗

arXiv2026

PulseQuant: Propagation-Guided Subspace Correction for 4-Bit Video Diffusion Transformers

Quantization errors in video diffusion transformers can be amplified or attenuated by subsequent denoising updates, making local reconstruction error an incomplete predictor of final impact. We introduce PulseQuant, a 4-bit post-training quantization method that combines trajectory sensitivity with activation geometry to guide offline calibration. Isolated block--step interventions estimate propagation risk, which prioritizes sensitive trajectory states during row-radius selection. With these radii fixed, response-subspace correction uses neighboring-code edits to reduce residual components along dominant activation directions. Both stages preserve the original 4-bit weight representation. Controlled interventions show that short-horizon propagated error predicts final latent error more reliably than immediate block-output error, supporting calibration beyond local reconstruction objectives. Evaluations on Wan models, Self Forcing, and MiniMax-H3 demonstrate improvements in key consistency and dense-reference metrics while remaining competitive on other attributes across model scales and generation paradigms.

cs.CV↗

arXiv2026

Rethinking Where to Edit: Task-Aware Localization for Instruction-Based Image Editing

Instruction-based image editing (IIE) aims to modify an image according to a natural language instruction. Despite recent advances in diffusion transformers, existing methods often introduce unintended changes to regions unrelated to the requested edit. We attribute this limitation to the absence of an explicit mechanism for edit localization. Different editing operations (e.g., subject addition, removal, and replacement) induce distinct spatial patterns, yet existing IIE models typically perform localization in a task-agnostic manner. To address this limitation, we propose a training-free, task-aware edit localization framework that exploits the intrinsic source and target image streams of IIE models. For each image stream, we construct feature centroids from attention-based edit cues, and then partition tokens into edit and non-edit regions based on feature similarity. Observing that effective localization is inherently task-dependent, we introduce a unified mask construction strategy that selectively leverages the source and target streams according to the editing task. We also provide a systematic analysis of our underlying insights and design choices. Extensive experiments on EdiVal-Bench demonstrate that our framework consistently improves content consistency in non-edit regions while maintaining strong instruction-following performance on top of Step1X-Edit and Qwen-Image-Edit.

cs.CV↗

arXiv2026

AGM: Achievement-Grounded Memory for Closed-Loop Agents with Frozen VLA Policies

Frozen vision-language-action (VLA) policies offer broad manipulation skills but execute open-loop action chunks without tracking task progress, so the agent cannot reliably decide whether to continue, retry, or terminate. External memory is a natural remedy, yet we find it can be harmful when attempted actions are recorded as completed progress: transient execution failures become persistent task-state errors, and such a memory can underperform no progress memory at all. We propose Achievement-Grounded Memory (AGM), a lightweight closed-loop framework for frozen VLA policies. AGM represents a task as a static subgoal sequence with a dynamic progress pointer and advances the pointer on physically verified achievement rather than on attempts. Proprioceptive gripper-load cues decide when to verify; coherent point tracking verifies grasps, and language-conditioned cross-view comparison, read by a single trained 2.43M-parameter verification head, verifies placements. The policy, tracker, and encoder remain frozen, the head is the only trained component, and deployment needs no auxiliary vision-language model. On the RoboMME Counting benchmark, AGM reaches 100.0% on PickXTimes and 84.0% on BinFill, surpassing the strongest memory-augmented baseline by 7.7 points on the four-task average, and the gains carry over to a physical robot, where AGM reaches 100.0% and 82.0%. These results suggest that reliable embodied memory depends more on disciplined state updates than on memory capacity.

cs.RO↗

arXiv2026

Learning to Act under Visual Interruptions with Vision-Language-Action Models

Vision-language-action (VLA) models have demonstrated strong capabilities in robotic manipulation, but they are typically developed and evaluated with all camera streams available throughout task execution. When a camera stops delivering frames during task execution, the policy must continue acting without access to subsequent observations from the missing view. Despite its practical importance, how such interruptions affect closed-loop manipulation remains insufficiently understood. To investigate this problem, we introduce MAIL-Bench, a benchmark that evaluates visual interruptions with VLA models. By interrupting different cameras at multiple stages of each policy's successful reference trajectory, MAIL-Bench measures how well policies retain their capabilities when visual inputs become unavailable. Building on this benchmark, we propose MINT, which first trains VLA policies to remain functional under missing visual inputs. At inference time, MINT selectively supplements missing observations using optical-flow extrapolation or an action-conditioned world model, and withdraws predicted views when they become unreliable. Experiments on $π_{0.5}$ and GR00T N1.5 show that MINT significantly improves task success under camera loss over the original models. Experiments on AgiBot G2 further demonstrate the real-robot deployment under camera loss. The benchmark is available at https://minglejiang.github.io/Mail-Bench/

cs.RO↗

arXiv2026

BEAT: Rhythm-Elastic Alignment for Agentic Music-guided Movie Trailer Generation

Automatic movie trailer generation must select shots from a full-length film and synchronize them with background music. Existing methods either relegate music alignment to post-processing or enforce rigid one-to-one shot-music mappings, overlooking that professional editing rhythm is elastic: rapid cuts accompany high-energy passages while sustained shots span quieter bars. We introduce BEAT, a framework that addresses this gap with two core components: MuVA, a compact music-visual alignment encoder trained with Sinkhorn-regularized two-stage learning, and Bar-DP, an energy-adaptive dynamic programming algorithm that produces elastic many-to-one alignments following musical dynamics. These components are integrated into a five-phase agentic pipeline that grounds the core alignment in learned cross-modal features while coordinating higher-level creative decisions through structured text signals. To support comprehensive evaluation, we also introduce TrailerArena, a benchmark with 20+ metrics across four complementary dimensions. On TrailerArena, BEAT achieves state-of-the-art performance across shot selection, ordering, and perceptual quality, while producing fully composed trailers end-to-end.

cs.CV↗

arXiv2026

RA-SOD: Reliability-Aware RGB-T Salient Object Detection under Modality Degradation

RGB-Thermal (RGB-T) salient object detection leverages complementary cues from visible and thermal modalities to improve robustness in challenging environments. However, in real-world scenarios, the reliability of each modality is inherently unstable: RGB images degrade under low illumination, motion blur, and noise, while thermal imagery often suffers from contrast compression and sensor artifacts. Such degradation introduces unreliable perceptual evidence that can mislead cross-modal fusion and significantly deteriorate detection performance. To address this challenge, we propose RA-SOD, a reliability-aware RGB-T salient object detection framework that explicitly models modality reliability and integrates it into feature learning and cross-modal fusion. First, we introduce a reliability-conditioned representation that adaptively compensates degraded modality features while preserving structural cues. Second, an uncertainty-guided dual-stream refinement strategy progressively corrects cross-modal representations while suppressing unreliable evidence. Finally, we propose a pixel-wise modality competition mechanism that dynamically selects modality cues according to spatial reliability for fine-grained fusion. Extensive experiments on four benchmarks (VT821, VT1000, VT5000, and VT-IMAG) demonstrate that RA-SOD achieves state-of-the-art performance and exhibits strong robustness under severe modality degradation. Code and models are available at https://github.com/zaoxienian/RA-SOD.

cs.CV↗

arXiv2026

StellaVLA: In-Context Structured Demonstration for Generalizable Vision-Language-Action Models

Vision-Language-Action (VLA) models can follow instructions and manipulate objects, but their performance often collapses out of distribution (OOD), when the scene, viewpoint, or object differs from training. Adapting to each new situation typically requires collecting more data and fine-tuning. We present StellaVLA, a framework that instead adapts at test time by conditioning on a single retrieved demonstration. The key idea is to move beyond imitating what an expert did and instead convey why: an automated offline pipeline converts each raw trajectory into a structured demonstration, e.g., a task plan, sub-goal descriptions, and verbalized 3D motion, at zero human-annotation cost. Provided as in-context guidance, this structured demonstration lets the policy reason about the task rather than mimic a pixel trajectory, which also makes it transferable across embodiments (real-robot, human-hand, or XR demonstrations). A parallel dual-training design internalizes this reasoning during training through a joint action-and-language objective, while inference uses the action expert alone, preserving real-time, high-frequency control with no added latency. On the VLA-Arena leaderboard(Aug 1, 2026), StellaVLA ranks first with an overall score of 0.63, versus 0.44 and 0.22 for the strong prior models ($π_{0.5}$ and LingBot-VLA), and it further leads on LIBERO with 98.8% average success rate and LIBERO-Plus with 85.1% success rate. Our real-robot benchmark demonstrates that StellaVLA can use both human/robot demos and human-to-robot (XR) demos as in-context structured demonstration to help VLA model adapt to OOD tasks.

cs.RO↗

arXiv2026

LoSA: Near-Lossless Sparse Attention for Training-Free Video Diffusion Acceleration

Video diffusion transformers are costly to sample: every denoising step applies self-attention over a long 3D token sequence, a quadratic cost that dominates as resolution and duration grow. Sparse attention reduces this cost without retraining, but existing methods pursue aggressive sparsity, where further speedup costs disproportionately more attention fidelity. We target the opposite end of this trade-off: fix near-lossless fidelity by construction, and remove as much computation as this constraint permits. Two observations make this regime practical: roughly 40% of block interactions can be removed while retaining 99% of the attention mass, and the high-mass support remains stable across denoising steps. We propose LoSA, a training-free sparse-attention method that fixes a retained-mass threshold of 99% rather than a sparsity ratio: it measures exact block attention masses at one early dense step, keeps, for each head and query block, the smallest key/value block set meeting the threshold, and reuses the frozen block indices for all remaining steps. On Wan2.1-1.3B, LoSA alone gives a $1.36\times$ speedup with a 0.06-point VBench Overall drop. The benefit is largest under composition: combined with feature caching, LoSA reaches a $3.2\times$ speedup on HunyuanVideo at a 0.02-point drop, versus 0.32 points for the strongest sparse baseline at comparable speed. Across three video diffusion transformers and speedups up to $3.2\times$, LoSA consistently achieves the best training-free speed-quality trade-off.

cs.CV↗

arXiv2026

NashPG: A Policy Gradient Method with Iteratively Refined Regularization for Finding Nash Equilibria

Finding Nash equilibria in two-player zero-sum imperfect-information games remains a central challenge in multi-agent reinforcement learning. Recent multi-round regularization methods offer a promising direction, yet existing approaches either require full enumeration of the game tree or rely on non-policy-gradient inner solvers that underperform in practice, leaving a scalable policy-gradient-based solution open. In this paper, we propose a novel multi-round regularization procedure and show that it guarantees strictly monotonic reduction in Bregman divergence to Nash equilibria and eventual convergence to one in two-player zero-sum extensive-form games. Guided by this framework, we develop a practical algorithm, Nash Policy Gradient (NashPG), which places the regularization directly in the policy optimization objective and is implemented using standard policy gradient methods. Empirically, NashPG achieves comparable or lower exploitability than prior model-free methods on classic benchmark games and scales to large domains such as Battleship and No-Limit Texas Hold'em, where it attains higher average payoff in head-to-head play.

cs.LG↗

arXiv2026

Diversifying Personalized Research Ideation against AI-Induced Homogenization

AI-assisted research ideation has emerged as a promising paradigm for accelerating scientific discovery, with systems now capable of generating research directions conditioned on papers, topics, or lightweight researcher contexts. Yet current systems largely optimize individual suggestions in isolation. This leaves two blind spots. First, coarse researcher representations may elicit mainstream directions that appear broadly feasible, but lack sufficient researcher-specific grounding. Second, independent recommendations can concentrate a community's portfolio around recurring high-probability themes. To address these blind spots, we propose DivAlign, a four-stage pipeline for alignment-preserving de-homogenization. DivAlign extracts fine-grained researcher profiles, generates profile-conditioned candidate directions, scores them along three alignment dimensions (Executability, Comprehensibility, and Growth Potential), and surfaces researcher-local directions while reducing redundancy across the community portfolio. On a benchmark we construct from 95 AI researchers across five subfields, DivAlign reduces community-level redundancy while preserving researcher-direction fit. Compared with coarse single-shot ideation, it lowers average pairwise similarity from 0.331 to 0.294 and nearest-neighbor similarity from 0.704 to 0.608. Compared with the independent top-choice variant, DivAlign reduces nearest-neighbor similarity from 0.663 to 0.608 while retaining 99.9% of the researcher-direction fit score. Code and data are available at https://github.com/Ruixxxx/DivAlign.

cs.AI↗

arXiv2026

HiCI: Hierarchical Construction-Integration for Long-Context Attention

Long-context language modeling is commonly framed as a scalability challenge of token-level attention, yet local-to-global information structuring remains largely implicit in existing approaches. Drawing on cognitive theories of discourse comprehension, we propose HiCI (Hierarchical Construction--Integration), a hierarchical attention module that constructs segment-level representations, integrates them into a shared global context, and broadcasts both to condition segment-level attention. We validate HiCI through parameter-efficient adaptation of LLaMA-2 with only <5.5% additional parameters, extending context from 4K to 100K tokens (7B) and 64K tokens (13B). Across language modeling, retrieval, and instruction-following benchmarks, HiCI yields consistent improvements over strong baselines, including matching proprietary models on topic retrieval and surpassing GPT-3.5-Turbo-16K on code comprehension. These results demonstrate the effectiveness of explicit hierarchical structuring as an inductive bias for long-context modeling.

cs.CL↗