arXiv subjects

Yi Yang

Publications and source records attributed to Yi Yang.

At least 289 records · Page 16Linked to original sources

arXiv2025

RAG Meets Temporal Graphs: Time-Sensitive Modeling and Retrieval for Evolving Knowledge

Knowledge is inherently time-sensitive and continuously evolves over time. Although current Retrieval-Augmented Generation (RAG) systems enrich LLMs with external knowledge, they largely ignore this temporal nature. This raises two challenges for RAG. First, current RAG methods lack effective time-aware representations. Same facts of different time are difficult to distinguish with vector embeddings or conventional knowledge graphs. Second, most RAG evaluations assume a static corpus, leaving a blind spot regarding update costs and retrieval stability as knowledge evolves. To make RAG time-aware, we propose Temporal GraphRAG (TG-RAG), which models external corpora as a bi-level temporal graph consisting of a temporal knowledge graph with timestamped relations and a hierarchical time graph. Multi-granularity temporal summaries are generated for each time node to capture both key events and broader trends at that time. The design supports incremental updates by extracting new temporal facts from the incoming corpus and merging them into the existing graph. The temporal graph explicitly represents identical facts at different times as distinct edges to avoid ambiguity, and the time hierarchy graph allows only generating reports for new leaf time nodes and their ancestors, ensuring effective and efficient updates. During inference, TG-RAG dynamically retrieves a subgraph within the temporal and semantic scope of the query, enabling precise evidence gathering. Moreover, we introduce ECT-QA, a time-sensitive question-answering dataset featuring both specific and abstract queries, along with a comprehensive evaluation protocol designed to assess incremental update capabilities of RAG systems. Extensive experiments show that TG-RAG significantly outperforms existing baselines, demonstrating the effectiveness of our method in handling temporal knowledge and incremental updates.

cs.IR↗

arXiv2025

ManiAgent: An Agentic Framework for General Robotic Manipulation

While Vision-Language-Action (VLA) models have demonstrated impressive capabilities in robotic manipulation, their performance in complex reasoning and long-horizon task planning is limited by data scarcity and model capacity. To address this, we introduce ManiAgent, an agentic architecture for general manipulation tasks that achieves end-to-end output from task descriptions and environmental inputs to robotic manipulation actions. In this framework, multiple agents involve inter-agent communication to perform environmental perception, sub-task decomposition and action generation, enabling efficient handling of complex manipulation scenarios. Evaluations show ManiAgent achieves an 86.8% success rate on the SimplerEnv benchmark and 95.8% on real-world pick-and-place tasks, enabling efficient data collection that yields VLA models with performance comparable to those trained on human-annotated datasets. The project webpage is available at https://yi-yang929.github.io/ManiAgent/.

cs.RO↗

arXiv2025

Learning from Reference Answers: Versatile Language Model Alignment without Binary Human Preference Data

Large language models~(LLMs) are expected to be helpful, harmless, and honest. In different alignment scenarios, such as safety, confidence, and general preference alignment, binary preference data collection and reward modeling are resource-intensive but play a central role in transferring human preferences. In this work, we explore using the similarity between sampled generations and reference answers as a supplementary reward function for alignment. When unary reference answers are available, such similarity-based rewards can circumvent the need for binary preference data and explicit reward modeling. We introduce \textit{RefAlign}, a versatile REINFORCE-style alignment algorithm that does not rely on reward or reference models. RefAlign utilizes language generation evaluation metrics, such as BERTScore, between sampled generations and reference answers as surrogate rewards. Beyond general preference optimization, RefAlign can be naturally extended to diverse scenarios, including safety and confidence alignment, by combining similarity-based rewards with task-specific objectives. Across multiple scenarios, RefAlign achieves performance comparable to prior alignment methods while operating without binary preference data or reward models. The code is available at https://github.com/mzhaoshuai/RefAlign.

cs.CL↗

arXiv2025

Translution: Unifying Self-attention and Convolution for Adaptive and Relative Modeling

When modeling a given type of data, we consider it to involve two key aspects: 1) identifying relevant elements (e.g., image pixels or textual words) to a central element, as in a convolutional receptive field, or to a query element, as in self-attention, and 2) encoding these tokens effectively. Self-attention can adaptively identify these elements but relies on absolute positional embedding for structural representation learning. In contrast, convolution encodes elements in a relative manner, yet their fixed kernel size limits their ability to adaptively select the relevant elements. In this paper, we introduce Translution, an operation that unifies the adaptive identification capability of self-attention and the relative encoding advantage of convolution. However, this integration leads to a substantial increase in the number of parameters, exceeding most currently available computational resources. Therefore, we propose a lightweight variant of Translution, named α-Translution. Experiments on computer vision and natural language processing tasks show that Translution (including α-Translution) achieves superior accuracy compared to self-attention. The code is available at https://github.com/hehefan/Translution.

cs.LG↗

arXiv2025

Decentralized Multi-Robot Relative Navigation in Unknown, Structurally Constrained Environments under Limited Communication

Multi-robot navigation in unknown, structurally constrained, and GPS-denied environments presents a fundamental trade-off between global strategic foresight and local tactical agility, particularly under limited communication. Centralized methods achieve global optimality but suffer from high communication overhead, while distributed methods are efficient but lack the broader awareness to avoid deadlocks and topological traps. To address this, we propose a fully decentralized, hierarchical relative navigation framework that achieves both strategic foresight and tactical agility without a unified coordinate system. At the strategic layer, robots build and exchange lightweight topological maps upon opportunistic encounters. This process fosters an emergent global awareness, enabling the planning of efficient, trap-avoiding routes at an abstract level. This high-level plan then inspires the tactical layer, which operates on local metric information. Here, a sampling-based escape point strategy resolves dense spatio-temporal conflicts by generating dynamically feasible trajectories in real time, concurrently satisfying tight environmental and kinodynamic constraints. Extensive simulations and real-world experiments demonstrate that our system significantly outperforms in success rate and efficiency, especially in communication-limited environments with complex topological structures.

cs.RO↗

arXiv2025

Upscaling Tomographic Volumetric 3D Printing via Virtual Stitching of Coordinated Projections

Tomographic volumetric 3D printing offers layer free, rapid fabrication of objects with high design freedom, but is limited to relatively small curing volumes because of the optical constraints imposed by an assumed need for telecentricity. We present a method to virtually stitch multiple projections from different light sources to build a single workpiece. To avoid the built in requirement for telecentricity and thus the need for an index matching vat, projections are produced by a graphics processing unit accelerated raytracing solver. The method accounts for non ideal light propagation, including nontelecentricity, reflection, refraction, attenuation, as well as any output power mismatch among projectors.Tomographic volumetric 3D printing offers layer free, rapid fabrication of objects with high design freedom, but is limited to relatively small curing volumes because of the optical constraints imposed by an assumed need for telecentricity. We present a method to virtually stitch multiple projections from different light sources to build a single workpiece. To avoid the built in requirement for telecentricity and thus the need for an index matching vat, projections are produced by a graphics processing unit accelerated raytracing solver. The method accounts for non ideal light propagation, including nontelecentricity, reflection, refraction, attenuation, as well as any output power mismatch among projectors.

physics.optics↗

arXiv2025

SSR-ZSON: Zero-Shot Object Navigation via Spatial-Semantic Relations within a Hierarchical Exploration Framework

Zero-shot object navigation in unknown environments presents significant challenges, mainly due to two key limitations: insufficient semantic guidance leads to inefficient exploration, while limited spatial memory resulting from environmental structure causes entrapment in local regions. To address these issues, we propose SSR-ZSON, a spatial-semantic relative zero-shot object navigation method based on the TARE hierarchical exploration framework, integrating a viewpoint generation strategy balancing spatial coverage and semantic density with an LLM-based global guidance mechanism. The performance improvement of the proposed method is due to two key innovations. First, the viewpoint generation strategy prioritizes areas of high semantic density within traversable sub-regions to maximize spatial coverage and minimize invalid exploration. Second, coupled with an LLM-based global guidance mechanism, it assesses semantic associations to direct navigation toward high-value spaces, preventing local entrapment and ensuring efficient exploration. Deployed on hybrid Habitat-Gazebo simulations and physical platforms, SSR-ZSON achieves real-time operation and superior performance. On Matterport3D and Habitat-Matterport3D datasets, it improves the Success Rate(SR) by 18.5\% and 11.2\%, and the Success weighted by Path Length(SPL) by 0.181 and 0.140, respectively, over state-of-the-art methods.

cs.RO↗

arXiv2025

GLUE: Global-Local Unified Encoding for Imitation Learning via Key-Patch Tracking

In recent years, visual representation learning has gained widespread attention in robotic imitation learning. However, in complex Out-of-Distribution(OOD) settings characterized by clutter and occlusion, the attention of global visual representations can be diluted or interfered, leading to degraded policy performance. The invariance of local representations for task-relevant objects offers a solution. By efficiently utilizing these local representations, training and testing data can be mapped to a more similar feature space, thereby mitigating the covariate shift problem. Accordingly, we propose GLUE, a global-local unified encoding framework for imitation learning based on key-patch tracking. GLUE selects and tracks key-patches as critical local representations by employing a text-guided mechanism. It features a novel fusion framework where global patch features query local patches to distill essential information, yielding fine-grained local features with low heterogeneity relative to the global context. This fused representation steers the robot's visual attention toward task-relevant objects and preserves precise global context, which together align the training and testing distributions into a similar and task-informative feature space, ultimately enhancing the robustness of the imitation learning policy. Experiments demonstrate that GLUE achieves strong performance across diverse tasks in both simulation and real-world settings, outperforming the strongest baseline by 17.6% in simulation, 36.3% in real-world environments, and 58.3% on real-world generalization settings. The project website of GLUE is available at https://GLUE666.github.io/.

cs.RO↗

arXiv2025

Dynamic Experts Search: Enhancing Reasoning in Mixture-of-Experts LLMs at Test Time

Test-Time Scaling (TTS) enhances the reasoning ability of large language models (LLMs) by allocating additional computation during inference. However, existing approaches primarily rely on output-level sampling while overlooking the role of model architecture. In mainstream Mixture-of-Experts (MoE) LLMs, we observe that varying the number of activated experts yields complementary solution sets with stable accuracy, revealing a new and underexplored source of diversity. Motivated by this observation, we propose Dynamic Experts Search (DES), a TTS strategy that elevates expert activation into a controllable dimension of the search space. DES integrates two key components: (1) Dynamic MoE, which enables direct control of expert counts during inference to generate diverse reasoning trajectories without additional cost; and (2) Expert Configuration Inheritance, which preserves consistent expert counts within a reasoning path while varying them across runs, thereby balancing stability and diversity throughout the search. Extensive experiments across MoE architectures, verifiers and reasoning benchmarks (i.e., math, code and knowledge) demonstrate that DES reliably outperforms TTS baselines, enhancing accuracy and stability without additional cost. These results highlight DES as a practical and scalable form of architecture-aware TTS, illustrating how structural flexibility in modern LLMs can advance reasoning.

cs.AI↗

arXiv2025

nnFilterMatch: A Unified Semi-Supervised Learning Framework with Uncertainty-Aware Pseudo-Label Filtering for Efficient Medical Segmentation

Semi-supervised learning (SSL) has emerged as a promising paradigm in medical image segmentation, offering competitive performance while substantially reducing the need for extensive manual annotation. When combined with active learning (AL), these strategies further minimize annotation burden by selectively incorporating the most informative samples. However, conventional SSL_AL hybrid approaches often rely on iterative and loop-based retraining cycles after each annotation round, incurring significant computational overhead and limiting scalability in clinical applications. In this study, we present a novel, annotation-efficient, and self-adaptive deep segmentation framework that integrates SSL with entropy-based pseudo-label filtering (FilterMatch), an AL-inspired mechanism, within the single-pass nnU-Net training segmentation framework (nnFilterMatch). By selectively excluding high-confidence pseudo-labels during training, our method circumvents the need for retraining loops while preserving the benefits of uncertainty-guided learning. We validate the proposed framework across multiple clinical segmentation benchmarks and demonstrate that it achieves performance comparable to or exceeding fully supervised models, even with only 5\%--20\% labeled data. This work introduces a scalable, end-to-end learning strategy for reducing annotation demands in medical image segmentation without compromising accuracy. Code is available here: https://github.com/Ordi117/nnFilterMatch.git.

cs.CV↗

arXiv2025

SparseDiT: Token Sparsification for Efficient Diffusion Transformer

Diffusion Transformers (DiT) are renowned for their impressive generative performance; however, they are significantly constrained by considerable computational costs due to the quadratic complexity in self-attention and the extensive sampling steps required. While advancements have been made in expediting the sampling process, the underlying architectural inefficiencies within DiT remain underexplored. We introduce SparseDiT, a novel framework that implements token sparsification across spatial and temporal dimensions to enhance computational efficiency while preserving generative quality. Spatially, SparseDiT employs a tri-segment architecture that allocates token density based on feature requirements at each layer: Poolingformer in the bottom layers for efficient global feature extraction, Sparse-Dense Token Modules (SDTM) in the middle layers to balance global context with local detail, and dense tokens in the top layers to refine high-frequency details. Temporally, SparseDiT dynamically modulates token density across denoising stages, progressively increasing token count as finer details emerge in later timesteps. This synergy between SparseDiT spatially adaptive architecture and its temporal pruning strategy enables a unified framework that balances efficiency and fidelity throughout the generation process. Our experiments demonstrate SparseDiT effectiveness, achieving a 55% reduction in FLOPs and a 175% improvement in inference speed on DiT-XL with similar FID score on 512x512 ImageNet, a 56% reduction in FLOPs across video generation datasets, and a 69% improvement in inference speed on PixArt-$α$ on text-to-image generation task with a 0.24 FID score decrease. SparseDiT provides a scalable solution for high-quality diffusion-based generation compatible with sampling optimization techniques.

cs.CV↗

arXiv2025

In-Context Edit: Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer

Instruction-based image editing enables precise modifications via natural language prompts, but existing methods face a precision-efficiency tradeoff: fine-tuning demands massive datasets (>10M) and computational resources, while training-free approaches suffer from weak instruction comprehension. We address this by proposing ICEdit, which leverages the inherent comprehension and generation abilities of large-scale Diffusion Transformers (DiTs) through three key innovations: (1) An in-context editing paradigm without architectural modifications; (2) Minimal parameter-efficient fine-tuning for quality improvement; (3) Early Filter Inference-Time Scaling, which uses VLMs to select high-quality noise samples for efficiency. Experiments show that ICEdit achieves state-of-the-art editing performance with only 0.1\% of the training data and 1\% trainable parameters compared to previous methods. Our approach establishes a new paradigm for balancing precision and efficiency in instructional image editing. Codes and demos can be found in https://river-zhang.github.io/ICEdit-gh-pages/.

cs.CV↗

arXiv2025

Text-Scene: A Scene-to-Language Parsing Framework for 3D Scene Understanding

Enabling agents to understand and interact with complex 3D scenes is a fundamental challenge for embodied artificial intelligence systems. While Multimodal Large Language Models (MLLMs) have achieved significant progress in 2D image understanding, extending such capabilities to 3D scenes remains difficult: 1) 3D environment involves richer concepts such as spatial relationships, affordances, physics, layout, and so on, 2) the absence of large-scale 3D vision-language datasets has posed a significant obstacle. In this paper, we introduce Text-Scene, a framework that automatically parses 3D scenes into textual descriptions for scene understanding. Given a 3D scene, our model identifies object attributes and spatial relationships, and then generates a coherent summary of the whole scene, bridging the gap between 3D observation and language without requiring human-in-the-loop intervention. By leveraging both geometric analysis and MLLMs, Text-Scene produces descriptions that are accurate, detailed, and human-interpretable, capturing object-level details and global-level context. Experimental results on benchmarks demonstrate that our textual parses can faithfully represent 3D scenes and benefit downstream tasks. To evaluate the reasoning capability of MLLMs, we present InPlan3D, a comprehensive benchmark for 3D task planning, consisting of 3174 long-term planning tasks across 636 indoor scenes. We emphasize clarity and accessibility in our approach, aiming to make 3D scene content understandable through language. Code and datasets will be released.

cs.CV↗

arXiv2025

Generalizability of Large Language Model-Based Agents: A Comprehensive Survey

Large Language Model (LLM)-based agents have emerged as a new paradigm that extends LLMs' capabilities beyond text generation to dynamic interaction with external environments. By integrating reasoning with perception, memory, and tool use, agents are increasingly deployed in diverse domains like web navigation and household robotics. A critical challenge, however, lies in ensuring agent generalizability - the ability to maintain consistent performance across varied instructions, tasks, environments, and domains, especially those beyond agents' fine-tuning data. Despite growing interest, the concept of generalizability in LLM-based agents remains underdefined, and systematic approaches to measure and improve it are lacking. In this survey, we provide the first comprehensive review of generalizability in LLM-based agents. We begin by emphasizing agent generalizability's importance by appealing to stakeholders and clarifying the boundaries of agent generalizability by situating it within a hierarchical domain-task ontology. We then review datasets, evaluation dimensions, and metrics, highlighting their limitations. Next, we categorize methods for improving generalizability into three groups: methods for the backbone LLM, for agent components, and for their interactions. Moreover, we introduce the distinction between generalizable frameworks and generalizable agents and outline how generalizable frameworks can be translated into agent-level generalizability. Finally, we identify critical challenges and future directions, including developing standardized frameworks, variance- and cost-based metrics, and approaches that integrate methodological innovations with architecture-level designs. By synthesizing progress and highlighting opportunities, this survey aims to establish a foundation for principled research on building LLM-based agents that generalize reliably across diverse applications.

cs.AI↗

arXiv2025

The stringy scaling loop expansion and stringy scaling violation

We propose a systematic approximation scheme to calculate general $n$-point $HSSA$ of open bosonic string theory. This stringy scaling loop expansion contains finite number of vacuum diagram terms at each loop order of scattering energy due to a vacuum diagram contraint and a topological graph constraint. In addition, we calculate coefficient and give the vacuum diagram representation and its Feynman rules for each term in the expansion of the $HSSA$. As an application to extending our previous calculation of $n$-point leading order stringy scaling behavior of $HSSA$, we explicitly calculate some examples of $4$-point next to leading order stringy scaling violation terms.

hep-th↗

arXiv2025

Ensemble of Pre-Trained Models for Long-Tailed Trajectory Prediction

This work explores the application of ensemble modeling to the multidimensional regression problem of trajectory prediction for vehicles in urban environments. As newer and bigger state-of-the-art prediction models for autonomous driving continue to emerge, an important open challenge is the problem of how to combine the strengths of these big models without the need for costly re-training. We show how, perhaps surprisingly, combining state-of-the-art deep learning models out-of-the-box (without retraining or fine-tuning) with a simple confidence-weighted average method can enhance the overall prediction. Indeed, while combining trajectory prediction models is not straightforward, this simple approach enhances performance by 10% over the best prediction model, especially in the long-tailed metrics. We show that this performance improvement holds on both the NuScenes and Argoverse datasets, and that these improvements are made across the dataset distribution. The code for our work is open source.

cs.LG↗

arXiv2025

Video2Roleplay: A Multimodal Dataset and Framework for Video-Guided Role-playing Agents

Role-playing agents (RPAs) have attracted growing interest for their ability to simulate immersive and interactive characters. However, existing approaches primarily focus on static role profiles, overlooking the dynamic perceptual abilities inherent to humans. To bridge this gap, we introduce the concept of dynamic role profiles by incorporating video modality into RPAs. To support this, we construct Role-playing-Video60k, a large-scale, high-quality dataset comprising 60k videos and 700k corresponding dialogues. Based on this dataset, we develop a comprehensive RPA framework that combines adaptive temporal sampling with both dynamic and static role profile representations. Specifically, the dynamic profile is created by adaptively sampling video frames and feeding them to the LLM in temporal order, while the static profile consists of (1) character dialogues from training videos during fine-tuning, and (2) a summary context from the input video during inference. This joint integration enables RPAs to generate greater responses. Furthermore, we propose a robust evaluation method covering eight metrics. Experimental results demonstrate the effectiveness of our framework, highlighting the importance of dynamic role profiles in developing RPAs.

cs.MM↗

arXiv2025

Image-based modelling of rock non-linear deformation under low-stress levels

Rock geophysical properties are widely reported to exhibit non-linear behaviours under low-stress conditions (below 10-20 MPa) before transitioning to the linear elastic stage, primarily due to the closure of microcracks and grain interfaces. Image-based modelling of rock deformation struggles to effectively characterise the microcrack closure effect because of the partial-volume effect, where image voxels are larger than microcracks and contain both pore and solid phases. This study presents a novel method to simulate non-linear rock deformation under elevated stress conditions. The method reconstructs digital rock models by treating partial-volume voxels as transitional phases that incorporate microcracks. By assigning intermediate elastic moduli and assuming that the pore portion within each partial-volume voxel deforms before the remaining solid content, the method employs the finite element method to simulate rock deformation and calculate the porosity of the deformed model. The method is tested on two Bentheimer sandstone models, and the results demonstrate its ability to predict the non-linear changes in porosity and elastic properties as the effective stress increases. This work provides a new pathway for image-based modelling of non-linear rock deformation considering the microcrack closure effect, offering valuable insights into the complex mechanical behaviour of rocks under confinement.

physics.geo-ph↗