Sparse
-
HINT-Blimp: Human INTent Inference from Multimodal Cues for Robotic Blimps
2026-09-22 23:38:19Subhadeep Koley, Benjamin Greenberg, Yifei Simon Shao, Juan Aceros, Nadia Figueroa, David Salda\~naAbstract
In human-robot interaction, traditional interfaces such as joysticks and handheld tablets introduce latency into navigation tasks and require the operator's explicit attention on the device, instead of the robot. We propose a new human-robot interaction framework in which a human communicates intent directly through sparse multimodal signals such as physical pushes and spoken commands. Human intent is represented as a parameterized linear dynamical system (LDS) that encodes the desired goal and motion behavior. The robot estimates this intent (parameters) online using a particle filter, where each particle represents a candidate LDS hypothesis and is reweighted online as new information becomes available. We validate this framework on a robotic blimp, whose inherent compliance and collision tolerance make it well-suited for repeated physical interaction. Experiments with multiple participants across 300 trials show that combining pushes and voice commands identifies the intended goal in 86% of trials within at most five interactions, with most trials resolved in two. The inferred dynamical systems can also produce curved trajectories that avoid obstacles known only to the human.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27154
PDF
-
Do We Need Complex Topology Control? Distinct-Peer Random Routing Improves Cost-Efficiency in Sparse Multi-Agent Debate
2026-09-22 23:25:43Boxuan Wang, Zhuoyun Li, Xiaowei Huang, Yi DongAbstract
Multi-agent debate (MAD) has emerged as a promising paradigm for improving the reasoning accuracy of large language models (LLMs) through iterative peer interaction. Communication topology plays a central role in this process, motivating increasingly sophisticated mechanisms that learn, adapt, or dynamically reconfigure agent interactions to improve accuracy or reasoning reliability. Meanwhile, prior studies suggest that much simpler sparse communication can already achieve competitive performance at substantially lower cost. In this work, we take a closer look at sparse MAD and ask whether complex topology control is actually necessary to improve collective reasoning. We find that a simple random-without-replacement routing policy, which lets each agent debate with two distinct and newly sampled peers at every round, provides a surprisingly strong baseline and consistently improves the accuracy-cost trade-off of sparse MAD. Building on this observation, we further study deliberation stopping and show that lightweight stopping can substantially reduce inference cost while preserving competitive accuracy. Our results suggest that sophisticated topology control such as learned topology adaption should be evaluated against strong simple routing and stopping baselines before its additional complexity is justified.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27150
PDF
-
PEARL: A Lightweight Prompt-based Feature Interpreter Framework for Real-Time, Anonymous, and Heterogeneous Collaborative Perception
2026-09-22 22:27:59Armin Maleki, Hayder RadhaAbstract
Heterogeneity across Collaborative Perception (CP) agents is a major challenge for emerging CP frameworks due to domain gaps from differing sensors, architectures, and training data. Prior works mitigate this challenge by aligning features in a unified space via model retraining or per-agent-type interpreters. These strategies (a) require access to neighbor configurations, (b) do not fully address real-time CP deployment, and (c) generalize poorly to unseen agents joining at run time. To overcome these challenges, we present PEARL, a Prompt-Embedding framework for Anonymous and Real-time Lightweight heterogeneous CP. PEARL supports multiple CP interpreters and selects one for a new-joining agent in real time using two lightweight, multi-scale interpreters trained in parallel: a sparse-detection (LWSD) interpreter that aligns salient regions for cooperative detection, and a dense, domain-invariant (LWDDI) interpreter that produces agent-invariant features for fast interpreter selection. Both interpreters use low-rank visual prompts to reduce computation, storage, and model complexity. Extensive experiments on simulated (OPV2V, V2XSet) and real (DAIR-V2X) datasets show that PEARL generalizes across simulated and real-world cooperative driving scenarios. Its real-time model-selection strategy yields an 8.2% Average Precision (AP) gain over a random-selection baseline while running in 1.67 ms on average. Although primarily designed for real-time CP, PEARL also outperforms state-of-the-art heterogeneous CP frameworks under traditional offline training by 5.6% AP on average while reducing communication cost by up to 34.7 times. Equally important, PEARL does not require sharing agents' configurations or model settings, thereby protecting information that may be proprietary or private. These results establish PEARL as a scalable and practical framework for heterogeneous collaborative perception.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27123
PDF
-
HiRE: Hindsight Reward Editing for Policy Finetuning
2026-09-22 21:04:10Haoyi Niu, Zhengtao Han, Yufeng Ji, Zhongyu Li, Koushil SreenathAbstract
Pre-trained robot policies always require finetuning to adapt to specific environments. Reinforcement Learning (RL) offers high performance potential because it improves action optimality rather than simply mimicking data. However, such potential depends heavily on reward quality. Sparse rewards lack process feedback, human-designed rewards are costly and biased, and semantic rewards from foundation representations are often not control-centric. We propose Hindsight Reward Editing (HiRE), a training-free framework to break this reward bottleneck. HiRE bridges the broad knowledge of foundation representation models with physical control awareness, by contrasting successful and failed trajectories in hindsight. It calibrates foundation representation models by identifying "trap states" that are predicted as high-rewarding states yet eventually result in failure, and vice versa. HiRE explicitly penalizes these traps while boosting rewards for critical successful states. This approach can be flexibly compatible with any foundation representations and RL algorithms. Experiments show that HiRE consistently outperforms other reward recipes by delivering dense, control-aware feedback that prevents value function collapse and reward hacking, thereby achieving superior sample efficiency, stable policy updates, and higher performance ceilings, e.g., at least 3x performance of the base policies. Qualitative results are at this https URL .
Abstract (translated)
URL
https://arxiv.org/abs/2609.27068
PDF
-
Reinforcement Learning with Decomposed Subtasks
2026-09-22 20:30:26Mattie Terzolo, Mikolaj Sacha, Ayan Sinha, Andrew RabinovichAbstract
Group Relative Policy Optimization (GRPO) and related policy-gradient methods for training language model agents collapse an entire multi-turn rollout into a single scalar trajectory reward before it enters the policy update. When the task composes distinct skills, especially under sparse and delayed environmental feedback, this collapsing is lossy: the optimizer must implicitly infer which competency drove the outcome and how that should change behavior. We argue the right primitive is not a better scalar but a decomposition: trajectory reward should be split along subtasks before it enters the policy update. We introduce Reinforcement Learning with Decomposed Subtasks (RLDS), whose core is Subtask-Decomposed Advantage Estimation (SDAE): a replacement for the scalar GRPO advantage that splits trajectory reward into per-subtask shares on a fixed taxonomy, computes a group-relative advantage per subtask, and distributes per-token credit by weighting each subtask's advantage by its importance, concentrating it around the step where a reflection marks that subtask's execution as consequential. We evaluate on four agentic benchmarks: FrozenLake (sparse grid navigation), HotpotQA (multi-hop QA, one retrieval tool), ScienceWorld (long-horizon embodied science), and DeepResearch (long-form research, four tools, composite rubric reward). Heterogeneity diagnostics emitted during training show where decomposition pays off - gains scale with subtask heterogeneity, largest on the high-heterogeneity tasks ScienceWorld (+11.5 points, paired-bootstrap 95% CI [+9.8, +13.3]) and FrozenLake (+9.8 points, [+7.0, +12.8]), and within noise on HotpotQA and DeepResearch, where the diagnostics predicted little to recover. ScienceWorld is also more compute-efficient under RLDS than scalar GRPO (-10.9% wall-clock per step), as long rollouts amortize the fixed reflect-and-grade overhead.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27035
PDF
-
Latent Commonality Expectation-Maximisation for Box-supervised Tree Crown Instance Segmentation
2026-09-22 15:05:31Thomas Pitts, Kunqi Li, Bin LiangAbstract
Individual tree crown segmentation from aerial imagery underpins tree-level carbon accounting, biodiversity, and restoration monitoring at landscape scale. However, existing models are predominantly trained on dense canopy forest imagery and degrade in savannah and drylands, where tree crowns are sparse, of variable appearance, and underrepresented in annotated benchmarks. These models also typically depend on costly polygon annotations. We introduce LACE (LAtent Commonality Expectation-maximisation), a box-supervised instance segmentation model, evaluated on 0.1 m/px aerial RGB tree crown imagery. LACE uses a frozen DINOv3-web ViT-L/16 encoder, applied at four spatial offsets and interlaced into a denser feature grid, with a lightweight CenterNet-style detection head trained solely on bounding boxes. We use expectation-maximisation to separate recurring appearance, the "treeness", within bounding boxes from surroundings. On the OAM-TCD benchmark test set, LACE reaches a mask AP$_{50}$ of $0.663 \pm 0.001$ (3 seeds) trained on 900 box-annotated images and without mask annotations, above the 0.626 scored by Restor's released mask-supervised Mask R-CNN, which was trained on the full ~4.2k image set. On a sparse-canopy holdout set, mask AP$_{50}$ rises to $0.691$ versus $0.612$ for Detectree2, a mask-supervised baseline. On NeonTreeEvaluation, using the official evaluation code, LACE reaches $0.728 \pm 0.003$ F1@0.4 (5 seeds) from 23,424 hand-annotated RGB boxes alone, matching the authors' DeepForest model's published 0.719, using under 0.1% of its training annotations and none of its LiDAR-derived 30M-crown pretraining set. By leveraging frozen self-supervised features, LACE matches or surpasses fully-supervised specialist baselines from boxes alone, removing the need for polygon annotation in tree crown instance segmentation for sparse-canopy environments where labelled data is scarce.
Abstract (translated)
URL
https://arxiv.org/abs/2609.26549
PDF
-
SparseNav: Instruction-conditioned Sparse Semantic Perception for Training-Free Vision-Language Navigation
2026-09-22 13:41:49Quanhua Chen, Juhan Kang, Runfeng Lin, ZiFei Zhang, Enquang Feng, Chunran Zheng, Xiwang Dong, Jiarong LinAbstract
Map-based vision-language navigation (VLN) relies on persistent spatial representations to connect language understanding with geometric planning. However, acquiring semantics beyond the needs of the current instruction can introduce unnecessary perception cost and irrelevant annotations. Continuously accumulating unrelated objects may not only waste computation, but also clutter the visual-spatial representation consumed by the vision-language model (VLM) planner. To address this problem, we present SparseNav, a training-free framework that follows a less-is-more principle for semantic navigation. SparseNav persistently maintains a lightweight geometric bird's-eye-view (BEV) map and sparse landmark memory, acquiring new semantics on demand using the active sub-instruction to decide what is worth grounding. An instruction manager first tracks navigation progress and identifies the active landmark query. An instruction-conditioned perception mechanism then invokes open-vocabulary segmentation when the queried landmark is visible and its metric location can inform the next decision. The resulting landmark memory supports VLM selection among hybrid frontier and local directional waypoint candidates. Without any additional training, SparseNav achieves success rates of 42.8% on R2R-CE and 40.7% on RxR-CE, both on the Val-Unseen splits. Controlled ablations examine semantic perception strategies and the contributions of individual framework components. Furthermore, we successfully deployed SparseNav on a Unitree Go2 quadruped equipped with an Intel RealSense D455 RGB-D camera for geometric mapping and landmark grounding and a Livox MID-360 LiDAR for localization, without a prebuilt map. We validated its effectiveness across multiple indoor environments using instruction-conditioned waypoint navigation.
Abstract (translated)
URL
https://arxiv.org/abs/2609.26408
PDF
-
HySparse2: Hybrid Sparse Attention with Two-Level KV Sharing
2026-09-22 13:10:32Jianyu Wei, Yizhao Gao, Qihao Zhang, Shimao Chen, Zhengju Tang, Yu Cheng, Shengjie Zhou, Zihan Jiang, Yifan Song, Hailin Zhang, Liang Zhao, Bo Yang, Gang Wang, Shijie Cao, Fuli LuoAbstract
Long-horizon and multi-turn agents typically generate short actions and process long observations from tools and environments. This growing context demands efficient prefill, compact KV-cache storage, and accurate long-context retrieval. To meet these demands, we introduce HySparse2, a hybrid sparse attention architecture with two-level KV sharing. At the outer level, KV Bridging adopts a YOCO-style self-decoder and cross-decoder structure, but bridges only full-attention layers. The self-decoder uses hybrid sliding-window attention (SWA), while the cross-decoder uses hybrid sparse attention. The KV caches for full-attention layers in the cross-decoder are generated from the hidden states of full-attention layers in the self-decoder. At the inner level, HySparse2 retains HySparse's core KV Reuse design with two refinements. First, it replaces block-level sparsity with token-level sparsity for finer long-context retrieval. Second, it removes the separate SWA branch from sparse layers and instead forces a sliding window of recent tokens into the sparse selection. This two-level KV sharing allows all cross-decoder KV caches to be constructed from self-decoder hidden states. Prefill can therefore exit after the self-decoder, skipping all cross-decoder layers. On an 80B-A3B MoE model, HySparse2 outperforms HySparse and Hybrid SWA on long-context retrieval and multi-turn agentic tasks, while substantially reducing prefill computation and KV-cache storage.
Abstract (translated)
URL
https://arxiv.org/abs/2609.26368
PDF
-
Leveraging Vision-Based Point Cloud Map Priors for Camera-Based 3D Object Detection and Online Vectorized HD Mapping
2026-09-22 12:36:48Markus K\"appeler, Rohit Mohan, Abhinav ValadaAbstract
Camera-based 3D object detection and online vectorized HD mapping provide compact scene representations for autonomous driving, but both depend on accurate metric geometry and remain limited by depth ambiguity. Over long-term deployment, observations from repeated traversals can be accumulated into persistent point cloud priors that provide geometric context beyond the current observations. Existing explicit point cloud prior approaches, however, rely on LiDAR-based map construction and therefore require expensive 3D ranging sensors. We propose a framework that constructs a static point cloud prior map from previous camera traversals using Pi3X and augments each point with DINOv3 features. At runtime, a local prior patch is retrieved using global localization, encoded with a sparse voxel backbone, and fused in bird's-eye view (BEV) with lifted multi-view camera features. Task-specific sparse transformer heads then predict 3D objects and vectorized map elements from the fused representation. On Argoverse 2, the vision-based prior improves a strong baseline from 0.287 to 0.299 CDS and from 0.669 to 0.750 vectorized mapping mAP. Ablations show that semantic DINOv3 features are particularly important for vectorized mapping. These results demonstrate that vision-built geometric-semantic priors provide an effective form of long-term scene memory for camera-based perception, improving both tasks without LiDAR for prior-map construction or online inference.
Abstract (translated)
URL
https://arxiv.org/abs/2609.26325
PDF
-
COPE: Continual Personalization of LLMs under Sparse User Feedback via User Embeddings and Self-Evaluation
2026-09-22 12:23:22Ruike Cao, Fugen Yao, Liang Dong, Jian Xu, Guanjun Jiang, Li XiaoAbstract
While Large Language Models (LLMs) have achieved remarkable results across various benchmarks, their alignment with normative values often results in homogenized responses that fail to address diverse user preferences. Existing training-free methods often occupy valuable context windows through prompt engineering, while training-based methods typically remain static post-training, failing to support the continual optimization required in real-world settings. To address these challenges, we propose COPE (Continual Optimization with Personalized embedding and self-Evaluation), a novel optimization framework tailored for real-world-motivated interaction settings with sparse user feedback. Our framework assigns learnable personalized embeddings to each user and synergistically integrates preference capture, self-evaluation calibration, and personalized response optimization within a single update step. A key innovation of our method is the use of self-evaluation to generate proxy rewards, enabling continuous model updates even when explicit user feedback is unavailable. Experiments show that COPE consistently outperforms strong training-free and training-based baselines under sparse feedback, and remains complementary to Retrieval-Augmented Prompting (RAP). Further analyses confirm COPE's reliable self-evaluation, meaningful preference patterns, stable general capabilities, and robustness under shifting preferences and alternative evaluators.
Abstract (translated)
URL
https://arxiv.org/abs/2609.26853
PDF
-
Predict Before You Step: Auditable Occupancy Forecasting for Dynamic Obstacle Avoidance under Sparse Guidance
2026-09-22 10:25:51Yuhui Mao, Fen Liu, Shenghai Yuan, Tianxin Hu, Ruimeng Liu, Rong SuAbstract
Legged robots under sparse waypoint guidance must avoid moving obstacles using partial, rapidly changing LiDAR observations. We present LOOP (Latent-recurrent Occupancy rollOut Policy), a local avoidance policy that connects sparse waypoint guidance to a frozen locomotion controller at 50 Hz. From occupancy and ego-velocity histories, a recurrent predictor forecasts future occupancy over a 1 s horizon by warping the current map with learned flow and visibility gates. These maps guide velocity selection through map-derived features and geometric risk estimates, providing an explicit interface for inspecting and replacing predictions. In encounter-synchronised Isaac Lab evaluations, LOOP achieves 57.1% head-on success at obstacle speeds of 2.5-3.2 m/s, exceeding a retrained reactive baseline by 8.2 percentage points. Comparisons with a rollout-free BEV policy show smaller, scenario-dependent gains from the prediction branch, including improved crossing success and reduced variability across training seeds at the highest head-on speeds. The adapter runs onboard a Unitree Go2 in 14.5 ms per step and completes all 16 real-world crossing trials without collision, demonstrating deployment feasibility.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25969
PDF
-
MatchFusion: Explicit-Implicit Instance Matching for Spatio-Temporal Multimodal Autonomous Driving
2026-09-22 08:24:12Xiaoyu Li, Jiajia Fu, Long Shi, Tianyu Du, Ruihang Li, Xian Wu, Lijun Zhao, Yingtao Zhang, Lining Sun, Ruifeng LiAbstract
Sparse instance representations provide a compact interface for spatial LiDAR-camera and temporal past-current interaction in multimodal perception and E2EAD. Effective interaction requires reliable instance correspondences despite geometric discrepancies and heterogeneous semantic representations. Attention-based methods exploit contextual semantics but often require specialized representation alignment, increasing computational overhead. In contrast, association based on structured object states is efficient and interpretable but lacks contextual evidence to resolve ambiguous matches. To combine these complementary strengths, we propose MatchFusion, a learnable instance matching and fusion module for spatio-temporal multimodal autonomous driving. MatchFusion initializes pairwise affinities using geometric similarity and category consistency, then selectively refines structurally plausible associations using instance embeddings. The resulting soft matchmap guides a common residual aggregation operator for adaptive information exchange. This unified matching-fusion formulation supports spatial LiDAR-camera and temporal past-current interaction, using multi-view image-plane geometry and motion-compensated BEV geometry as the respective structural priors. Experiments on nuScenes demonstrate consistent perception gains across diverse front-end configurations. Compared with a prior instance-centric fusion method, the MatchFusion-equipped system achieves higher perception accuracy while reducing FLOPs by 55.3% and GPU memory usage by 39.3%, with the matching-fusion module accounting for only 3.7% of total perception latency. Integrating temporal MatchFusion into SparseDrive further improves perception within an E2E framework without additional supervision. These results establish explicit-implicit matching as an effective and efficient mechanism for spatio-temporal instance interaction.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25860
PDF
-
LiFR v2: Completion-Augmented Event Propagation for High-Rate Dense Prediction
2026-09-22 07:36:01Tao Wan, Xiaoshan Wu, Yifei Yu, Bo Wang, Xiaoyang Lyu, Muxin Liu, Aoxuan Pan, Zhongrui Wang, Xiaojuan QiAbstract
High-rate dense perception in dynamic environments is limited by the low update rate of RGB cameras, as rapid scene changes can occur between frames. Event cameras offer temporally dense but spatially sparse measurements, complementary to spatially dense RGB observations. Direct fusion cannot fully exploit this complementarity, while event-guided propagation fails on newly appearing or disoccluded regions without valid RGB support. We present LiFR v2, a unified propagation-completion-memory framework for causal anytime and streaming dense prediction from an RGB keyframe and events. LiFR v2 introduces an Event-Guided Completion Module (EGCM) to recover task-relevant representations where propagation is unsupported, and a History Retrieval Module (HRM) to reuse completed representations across successive queries. The framework supports semantic segmentation, monocular depth estimation, and multi-task dense prediction, and we further introduce SHF-Emerge to evaluate rapid object emergence and disocclusion. LiFR v2 achieves 74.37% mIoU on DSEC and 56.13% on SHF-Emerge, improving LiFR-Seg by 1.85 percentage points on the latter, while reducing SHF-Emerge depth RMSE from 1.564 m to 1.118 m over the propagation baseline. It also exceeds 100 FPS for both segmentation and depth, demonstrating accurate and efficient high-rate perception beyond RGB frame rates.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25803
PDF
-
PLAT: Sparse Timed Keyframe Motion Tracking for Humanoid Control via Privileged Latent Transition Learning
2026-09-22 06:47:41Zepeng Wang, Jiangxing Wang, Chao Ma, Xiaochuan Shi, Zongqing LuAbstract
Humanoid motion tracking policies rely on dense frame-by-frame references, limiting their use as high-level motion controllers for planning and interactive motion generation. We study \emph{Sparse Timed Keyframe Motion Tracking}, where a policy receives only sparse future keyframes and their desired arrival times, and must execute stable whole-body motions that reach successive goals. We propose \textbf{PLAT}, a three-stage sparse timed keyframe motion tracking policy learning framework with \textbf{P}rivileged \textbf{LA}tent \textbf{T}ransition learning. PLAT bridges dense motion tracking and sparse goal-conditioned control by exploiting dense goal sequences as privileged supervision during training while requiring only sparse timed keyframe commands at deployment. A pretrained dense tracking expert first provides robust motion priors. A privileged latent prior is then learned through DAgger-style imitation, followed by latent residual reinforcement learning that refines latent transitions instead of directly optimizing actions. Extensive simulation experiments demonstrate that PLAT maintains accurate and stable sparse timed keyframe tracking across varying planning horizons, with particularly strong performance under long-horizon commands. Successful deployment on a Unitree G1 humanoid robot further demonstrates the effectiveness and practicality of PLAT for sparse humanoid motion control.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25754
PDF
-
Parameterized Dense-Sparse Fusion for Hybrid Retrieval: Tuning a Rank-Score Mix on BEIR SciFact with Qdrant
Abstract
We study a parameterized hybrid ranker that fuses a dense embedding list and a sparse lexical list. The method has a small, explicit parameter vector: a dense prior $\alpha \in [0,1]$, a score-versus-rank mix $\lambda \in [0,1]$, an RRF smoothing parameter $\kappa > 0$, optional list-geometry coefficients that move $\alpha$ per query, and a router margin $\tau$ that can turn sparse search off. We grid-search those ranges on SciFact train (809 queries) and freeze the chosen values on SciFact test (300). The tuned rank-score mix ($\alpha = 0.8$, $\lambda = 0.75$, $\kappa = 20$) reaches 0.753 nDCG@10 and 0.889 recall@10, outperforming dense BGE (0.742 / 0.871) and equal-weight RRF (0.707 nDCG@10) on that test split. A list-conditioned $\alpha$ adds +0.0006 nDCG; a sparse-off router is rejected by the same train split (any $\tau$ that skipped approximately 50% of queries lost nDCG). These coefficients are dataset-specific. Equal RRF with the same models does not beat dense on a nine-zip BEIR macro-average (0.479 vs. 0.519 nDCG@10). Repeating the same train-then-freeze sweep independently on all 20 indexed units beats equal RRF on 20/20 and dense on 16/20 (unit-mean nDCG@10 0.467 vs. 0.462 dense vs. 0.420 RRF). Other corpora should reuse the ranges, not a copy of the SciFact point.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22770
PDF
-
Qwen3.8-Omni: Towards Native Omni-Modal Agents
Abstract
We introduce Qwen3.8-Omni-Flash, a natively multimodal agentic model for real-world multimodal productivity. Compared with previous omni models, which primarily emphasized perception and interaction, Qwen3.8-Omni-Flash substantially improves multimodal understanding and reasoning, as well as performance on long-horizon agentic tasks. These capabilities are supported by a native multimodal co-training strategy that preserves strong text-domain capabilities while facilitating the transfer of agentic capabilities from text to audio and video tasks. The model inherits the sparse mixture-of-experts (MoE) architecture of Qwen3.8-Next and extends the context window to one million tokens, supporting long-context multimodal reasoning and long-horizon planning. These advances enable integration into production workflows as a primary agent or a specialized sub-agent, supporting video editing, long-form audio and video translation, music-conditioned music video or movie generation, and video-based note or omni-skill creation. To address the lack of native audio and video support in existing agent harnesses, we release Qwen-MM-Plugins, a lightweight open-source plugin framework for multimodal productivity. We further frame real-time multimodal interaction as a system-level challenge requiring orchestration of context and memory management, tool use, and sub-agent delegation. Accordingly, we release Qwen-Live-Harness, an open-source framework for building responsive, real-time multimodal agents based on Qwen3.8-Omni-Flash. Extensive evaluations demonstrate that Qwen3.8-Omni-Flash achieves strong performance across multimodal understanding, reasoning, long-horizon agentic execution, and video productivity tasks. These results and the accompanying open-source tools support Qwen3.8-Omni-Flash as a practical foundation for deploying natively multimodal agents in research and production.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25611
PDF
-
Hi-OPD: Hierarchy-Aware Open-Prompt Detection for Remote Sensing Images
2026-09-22 02:30:42Jinlong Hu, Yi Zhang, Zhiqi Xia, Yikang Zhou, Shunping JiAbstract
Hi-OPD addresses a failure mode left uncontrolled by flat open-prompt training: descendant retrieval need not persist under ancestor queries when multi-source remote sensing annotations exhibit inconsistent granularity and missing labels. A detector may localize \textit{car} and \textit{van} under atomic prompts yet miss the same instances under \textit{vehicle}; flat AP does not expose this cross-level inconsistency. We propose Hi-OPD, a hierarchy-aware open-prompt detector, and construct RS153-HierOPD from 175,644 retained training image/tile records and 3.48M boxes mapped to 153 atomic categories with sparse hierarchy and alias relations. Hi-OPD learns ancestor retrieval through hierarchy-safe negative sampling, path multi-positive supervision, and one-way upward consistency, while per-source risk exclusion handles potentially missing labels. ConvVPE converts K-shot support boxes into text-compatible embeddings using detector-native features and the shared contrastive head. On Track A, Hi-OPD obtains 79.7/72.3 AP50 on DIOR/DOTA-v2.0, above the literature-reported OpenRSD results of 76.7/71.8. Under controlled training on the original converted annotations, the full hierarchy recipe raises DOTA-v2.0 parent AP50 from 7.2 to 71.5 and FAIR1M grandparent AP50 from 31.6 to 71.4, while DOTA-v2.0 atomic AP50 changes from 71.4 to 72.3. The text path reaches 99.7% CAR50 (0.3% violation) across the three common sources and 99.9%/0.1% on FAIR1M grandparent relations. On held-out VEDAI, text AP50 is 75.9, 6.2 points above OpenRSD. Joint AP and CAR show that explicit hierarchy training repairs this failure mode while retaining atomic detection and prompt transfer.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25584
PDF
-
Point Diffusion Mamba: Unified Diffusion-State-Space Modeling for Single-View 3D Reconstruction under Data Scarcity
2026-09-22 01:11:24Wei Zhou, Xinzhe Shi, Xingxing Hao, Xing Hao, Kang Li, Jinye Peng, Ying HeAbstract
While single-view 3D reconstruction has seen significant progress, extrapolating complex 3D structures from inherently ambiguous 2D observations remains fundamentally ill-posed, particularly in the critically underexplored data-scarce regime. To address this challenge, we propose Point Diffusion Mamba (PDM), a method that integrates the generative power of diffusion models with the efficiency of state-space model for single-view 3D reconstruction under data-scarce conditions. Specifically, PDM employs a lightweight reconstruction module tailored to handle unordered point-cloud inputs effectively. By combining a Local Geometric Aggregation module with Mamba blocks, our approach jointly models global geometric structures and local details. In 3D reconstruction, each point in the initial noisy input requires a precise prediction, yet the high-level features extracted by the Mamba module capture only abstract semantic information from sparse points. To bridge this gap, we introduce the Hierarchical Feature Integration Network, which fuses high-level semantic and local geometric features for each point, overcoming the limitations of token-based point-cloud reconstruction. Furthermore, we propose a Dynamic Weighted Sampling strategy that adaptively unifies 3D generation with single-view reconstruction by leveraging generative priors to enhance reconstruction quality. Experimental results on the ShapeNet and Pix3D benchmarks demonstrate that PDM outperforms state-of-the-art methods, providing an effective solution for 3D reconstruction under data-scarce settings. Code is available at: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25538
PDF
-
Matryoshka attribution: Learning to attribute language model outputs to representations and weights
2026-09-22 00:23:52Aryaman Arora, Kirill Acharya, Nathan Hu, Yanzhe Zhang, Noah Goodman, Dan Jurafsky, Christopher PottsAbstract
Attributing language model outputs to their internal computations is an open problem in interpretability. Existing methods, which use causal interventions, gradients, or learnable masks, either are infeasibly expensive or struggle to identify actual causally-important internal computations. We propose framing attribution as the problem of identifying nested subsets of internal components which minimise a downstream loss. To learn this task, we introduce Matryoshka Attribution (MAttr), a mask learning method that parametrises the mask with a simple differentiable sigmoid top-$k$ operator. We supervise training over all sparsities simultaneously by randomising $k$ over training, resulting in a learned ordering of components by attribution score. MAttr achieves number 1 on the official leaderboard of the Mechanistic Interpretability Benchmark (Mueller et al., 2025); our method identifies sparse and task-transferrable circuits across varying circuit bases. As a practical application, we show that MAttr can be trained with reinforcement learning to identify weight changes responsible for downstream behaviours in LLM finetuning. We train MAttr on refusal judge scores and find that restoring $1\%$ of Llama 3.1 8B Instruct's weights to their base model state is sufficient to remove refusals while maintaining capabilities. We view MAttr as a successful formulation of interpretability into a learnable objective that we can tackle with gradient descent, and encourage future work along these lines.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25518
PDF
-
SBMVTrack: Spike-Budgeted Multi-View Learning for Energy-Efficient UAV Tracking
2026-09-21 23:59:34Pengzhi Zhong, Jiwei Mo, Haolun Li, Ge Zheng, Jingqi Wang, Xinyi Bo, Shuiwang LiAbstract
With sparse and event-driven computation, spiking neural networks show great potential for achieving accurate and energy-efficient UAV visual tracking. However, existing SNN-based trackers typically use spike firing rates only for energy evaluation and lack explicit optimization of actual spike activity. To address this, we propose SBMVTrack, a fully spiking framework for energy-efficient UAV tracking. SBMVTrack introduces Energy-Weighted Spike Budgeting (EWSB). EWSB weights actual spike activity according to the computational cost of each spiking layer. It constrains the energy-weighted firing rate and saturation activity, thereby reducing redundant spike computations. To improve tracking performance under the spike budget constraint, we propose Masked Multi-View Target Modeling (MVTM). This method treats the initial template, online template, and search region from the same sequence as correlated temporal views. It enhances the robustness of target representations through cross-view feature completion and identity-consistency learning. Extensive experiments on multiple benchmarks demonstrate that SBMVTrack effectively reduces the average spike firing rate and theoretical energy consumption. Meanwhile, it maintains competitive tracking performance, achieving a better accuracy-energy trade-off. The source code will be released upon acceptance.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25503
PDF
