arXiv subjects

Shiyuan Deng

Publications and source records attributed to Shiyuan Deng.

8 recordsLinked to original sources

arXiv2026

Robo-Harness K1: Harnessing Robot-Use Agents via Perception Augmentation

Foundation vision-language models (VLMs) understand objects, instructions, and spatial relations, yet translating this capability into robotic manipulation remains difficult. Vision-language-action (VLA) models require extensive demonstrations and may compromise pretrained understanding, while direct RGB-only VLM control is costly and strongly dependent on model capability. We introduce Robo-Harness K1, a robot-use agent (RUA) framework that exposes perception as tools. The agent queries calibrated depth, persistent visual anchors, spatial measurements, and grasp hypotheses, then selects generic motions from the returned evidence. This interface makes 3D geometry accessible without changing the VLM architecture or training a depth encoder. On matched LIBERO-PRO tasks, Gemini 3.7 Flash with K1 reaches 77.8% accuracy, surpassing GPT-6 Astra's 61.1% with an RGB-only harness; K1 further improves Astra to 88.9%. Without target fine-tuning, Gemini with K1 transfers to three RoboSuite arms and dual-arm RoboTwin tasks. On RoboTwin, it achieves 32.0% on Easy and 28.0% on Hard, showing resilience to visual and environmental perturbations. K1 also produces tool-call traces aligned with next-token training. A Qwen3.5-9B student trained on only 107 teacher episodes reaches 44.2% accuracy on new initial states versus 30.2% for OpenVLA, and 13.9% on held-out task conditions versus 0.0% for OpenVLA. These results suggest that perception-augmented RUAs offer a promising route to sample-efficient, generalizable robotic policies that leverage VLM capabilities through an accessible tool interface.

cs.RO↗

arXiv2026

World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal

General-purpose vision-language models (VLMs) bring broad knowledge and spatial reasoning to robot manipulation, yet existing systems either use them indirectly, to predict constraints or write programs, or give them a view of the scene rather than a world in which to act. We present World Action Agent (WAA), a multi-agent harness through which VLMs pilot robots with basic tools, making every decision within a visual action workspace. The workspace has three properties. Contact views, selected automatically from the scene geometry, present the scene around the current interaction. Action rehearsal turns each action into an editable proposal that the agent, alone or through an Imagination Agent, previews and revises against planning feedback before execution. In-view correction closes the loop between observation, rehearsal, and low-level execution, letting the agent remove residual offsets in the view where it observes them. Through the same workspace, WAA acquires embodied procedural knowledge in two ways: it evolves multimodal skills from expert videos and human teaching under evidence-based review and consults them through a Skill Agent, and its interaction traces train smaller VLMs to pilot the same harness. On LIBERO-Pro, WAA with skills evolved only from LIBERO-90 reaches a state-of-the-art 75.6% average success, outperforming end-to-end VLAs, code-as-policy agents, and a visual-harness baseline with the same backbone; the same skills remain effective on robosuite without further learning. Fine-tuning Qwen3.5-9B on harness traces raises its out-of-domain success from 1.7% to 43.3%.

cs.RO↗

arXiv2026

CAM: Question Answering on Entity-Centric Videos with Continuous Extraction and Adaptive Querying

Memory facilitates question answering over long videos by extracting and retrieving facts to fit within the limited context windows of multimodal LLMs (MLLMs). Existing solutions typically extract independent memory entries from fixed-length video clips and thus cannot capture high-level semantics that need to be summarized over extended time periods, such as character traits and relations. Moreover, they rely solely on similarity-based retrieval and may fail to retrieve the fine-grained details required for question answering. To tackle these problems, we propose CAM, featuring continuous extraction for high-level semantics and adaptive querying for fine-grained details. In particular, CAM stores the entities and relations extracted from video clips in a knowledge graph. To capture the high-level semantics of each entity or relation, CAM summarizes the local subgraph of the target entity or relation once the subgraph reaches a predefined size. To retrieve the fine-grained details required for question answering, CAM supports multiple search methods, including knowledge graph traversal, video re-watching, and audio listening. It utilizes a planner-executor-verifier pipeline to adaptively compose these search methods according to question intent. Evaluations on three benchmarks show that CAM outperforms SOTA baselines and improves their accuracy by up to 23 percentage points. Code is available at https://github.com/Jake-Tian/CAM.

cs.CV↗

arXiv2026

RecMem: Recurrence-based Memory Consolidation for Efficient and Effective Long-Running LLM Agents

Memory systems often organize user-agent interactions as retrievable external memory and are crucial for long-running agents by overcoming the limited context windows of LLMs. However, existing memory systems invoke LLMs to process every incoming interaction for memory extraction, and such an eager memory consolidation scheme leads to substantial token consumption. To tackle this problem, we propose RecMem by rethinking when memory consolidation should be conducted. RecMem stores incoming interactions in a subconscious memory layer and encode them using lightweight embedding models for retrieval. LLMs are only invoked to extract episodic and semantic memory when sustained recurrence are observed for semantically similar interactions. Such recurrence-based consolidation works because these interactions correspond to a semantic cluster with rich information and thus are worth extraction and summarization. To improve accuracy, RecMem also incorporates a semantic refinement mechanism that recovers the fine-grained facts omitted by memory extraction. Experiments show that RecMem reduces the memory construction token cost of three SOTA memory systems by up to 87% while exceeding their accuracy.

cs.CL↗

arXiv2026

SAQ: Pushing the Limits of Vector Quantization through Code Adjustment and Dimension Segmentation

Approximate Nearest Neighbor Search (ANNS) plays a critical role in applications such as search engines, recommender systems, and RAG for LLMs. Vector quantization (VQ), a crucial technique for ANNS, is commonly used to reduce space overhead and accelerate distance computations. However, despite significant research advances, state-of-the-art VQ methods still face challenges in balancing encoding efficiency and quantization accuracy. To address these limitations, we propose a novel VQ method called SAQ. To improve accuracy, SAQ employs a new dimension segmentation technique to strategically partition PCA-projected vectors into segments along their dimensions. By prioritizing leading dimension segments with larger magnitudes, SAQ allocates more bits to high-impact segments, optimizing the use of the available space quota. An efficient dynamic programming algorithm is developed to optimize dimension segmentation and bit allocation, ensuring minimal quantization error. To speed up vector encoding, SAQ devises a code adjustment technique to first quantize each dimension independently and then progressively refine quantized vectors using a coordinate-descent-like approach to avoid exhaustive enumeration. Extensive experiments demonstrate SAQ's superiority over classical methods (e.g., PQ, PCA) and recent state-of-the-art approaches (e.g., LVQ, Extended RabitQ). SAQ achieves up to 80% reduction in quantization error and accelerates encoding speed by over 80x compared to Extended RabitQ.

cs.DB↗

arXiv2025

On estimation and order selection for multivariate extremes via clustering

We investigate the estimation of multivariate extreme models with a discrete spectral measure using spherical clustering techniques. The primary contribution involves devising a method for selecting the order, that is, the number of clusters. The method consistently identifies the true order, i.e., the number of spectral atoms, and enjoys intuitive implementation in practice. Specifically, we introduce an extra penalty term to the well-known simplified average silhouette width, which penalizes small cluster sizes and small dissimilarities between cluster centers. Consequently, we provide a consistent method for determining the order of a max-linear factor model, where a typical information-based approach is not viable. Our second contribution is a large-deviation-type analysis for estimating the discrete spectral measure through clustering methods, which serves as an assessment of the convergence quality of clustering-based estimation for multivariate extremes. Additionally, as a third contribution, we discuss how estimating the discrete measure can lead to parameter estimations of heavy-tailed factor models. We also present simulations and real-data studies that demonstrate order selection and factor model estimation.

stat.ME↗

arXiv2023

Space-Query Tradeoffs in Range Subgraph Counting and Listing

This paper initializes the study of {\em range subgraph counting} and {\em range subgraph listing}, both of which are motivated by the significant demands in practice to perform graph analytics on subgraphs pertinent to only selected, as opposed to all, vertices. In the first problem, there is an undirected graph $G$ where each vertex carries a real-valued attribute. Given an interval $q$ and a pattern $Q$, a query counts the number of occurrences of $Q$ in the subgraph of $G$ induced by the vertices whose attributes fall in $q$. The second problem has the same setup except that a query needs to enumerate (rather than count) those occurrences with a small delay. In both problems, our goal is to understand the tradeoff between {\em space usage} and {\em query cost}, or more specifically: (i) given a target on query efficiency, how much pre-computed information about $G$ must we store? (ii) Or conversely, given a budget on space usage, what is the best query time we can hope for? We establish a suite of upper- and lower-bound results on such tradeoffs for various query patterns.

cs.DS↗

arXiv2019

Pyramid: A General Framework for Distributed Similarity Search

Similarity search is a core component in various applications such as image matching, product recommendation and low-shot classification. However, single machine solutions are usually insufficient due to the large cardinality of modern datasets and stringent latency requirement of on-line query processing. We present Pyramid, a general and efficient framework for distributed similarity search. Pyramid supports search with popular similarity functions including Euclidean distance, angular distance and inner product. Different from existing distributed solutions that are based on KD-tree or locality sensitive hashing (LSH), Pyramid is based on Hierarchical Navigable Small World graph (HNSW), which is the state of the art similarity search algorithm on a single machine. To achieve high query processing throughput, Pyramid partitions a dataset into sub-datasets containing similar items for index building and assigns a query to only some of the sub-datasets for query processing. To provide the robustness required by production deployment, Pyramid also supports failure recovery and straggler mitigation. Pyramid offers a set of concise API such that users can easily use Pyramid without knowing the details of distributed execution. Experiments on large-scale datasets show that Pyramid produces quality results for similarity search, achieves high query processing throughput and is robust under node failure and straggler.

cs.DC↗