BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Jiahao Li

Jiahao Li contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Computer VisionArtificial IntelligenceComputation and Languageeess.IVHuman-Computer InteractionComputational GeometryGraphicsMachine LearningMultimediaphysics.app-phquant-phRobotics

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
15works
0followers
12topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

15 published item(s)

preprint2026arXiv

An Efficient Streaming Video Understanding Framework with Agentic Control

Streaming video requires handling dynamic information density under strict latency budgets. Yet, existing methods typically employ static strategies, such as fixed memory compression or reliance on a single model, forcing a trade-off: fast models fail on complex queries, while always-on heavy models violate real-time constraints and overcomplicate simple queries. Rather than fixing these decisions upfront, we propose R3-Streaming (Remember, Respond, Reason), which formulates streaming video understanding as a cascaded control problem: for each query, the system compresses memory, judges response readiness, and routes computation sequentially, so that each downstream decision builds on progressively refined information states. To optimize this pipeline, we introduce an age-aware forgetting policy for memory compression, as aggressively compressing historical frames can yield substantial performance gains. For compute routing, we propose TB-GRPO, a target-balanced reinforcement learning objective that routes hard queries to a stronger model while preventing mode collapse. Extensive evaluations demonstrate that R3-Streaming achieves state-of-the-art results among streaming MLLMs, reaching 57.92 on OVO-Bench and 76.36 on StreamingBench, while reducing visual token usage by 95 to 96 percent.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Efficient 3D Content Reconstruction and Generation

Automatic 3D content creation seeks to replace labor-intensive modeling and scanning pipelines with systems that can synthesize or recover 3D assets directly from text or images. Its applications span video games, virtual reality, robotics, and simulation, enabling rapid asset prototyping, diverse interactive world generation, and efficient 3D data collection for training foundation models. Contemporary solutions largely follow two complementary paradigms: (i) text- or image-to-3D generation, which learns priors over 3D geometry and appearance to create novel assets from natural language or a single view image; and (ii) 3D reconstruction, which estimates camera poses and geometry from RGB images. This thesis advances both directions. On the generation side, I introduce Instant3D, which combines multi-view diffusion with feed-forward sparse-view 3D reconstruction to produce high-quality assets in 5-20 seconds. On the reconstruction side, I develop FastMap, a structure-from-motion pipeline that achieves up to 10x speedup over prior state-of-the-art by using first-order optimization with fused GPU kernels extensively, while maintaining comparable pose accuracy and downstream novel view synthesis quality.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training

GUI agents that interact with graphical interfaces on behalf of users represent a promising direction for practical AI assistants. However, training such agents is hindered by the scarcity of suitable environments. We present InfiniteWeb, a system that automatically generates functional web environments at scale for GUI agent training. While LLMs perform well on generating a single webpage, building a realistic and functional website with many interconnected pages faces challenges. We address these challenges through unified specification, task-centric test-driven development, and a combination of website seed with reference design image to ensure diversity. Our system also generates verifiable task evaluators enabling dense reward signals for reinforcement learning. Experiments show that InfiniteWeb surpasses commercial coding agents at realistic website construction, and GUI agents trained on our generated environments achieve significant performance improvements on OSWorld and Online-Mind2Web, demonstrating the effectiveness of proposed system.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Qwen-Image-2.0 Technical Report

We present Qwen-Image-2.0, an omni-capable image generation foundation model that unifies high-fidelity generation and precise image editing within a single framework. Despite recent progress, existing models still struggle with ultra-long text rendering, multilingual typography, high-resolution photorealism, robust instruction following, and efficient deployment, especially in text-rich and compositionally complex scenarios. Qwen-Image-2.0 addresses these challenges by coupling Qwen3-VL as the condition encoder with a Multimodal Diffusion Transformer for joint condition-target modeling, supported by large-scale data curation and a customized multi-stage training pipeline. This enables strong multimodal understanding while preserving flexible generation and editing capabilities. The model supports instructions of up to 1K tokens for generating text-rich content such as slides, posters, infographics, and comics, while significantly improving multilingual text fidelity and typography. It also enhances photorealistic generation with richer details, more realistic textures, and coherent lighting, and follows complex prompts more reliably across diverse styles. Extensive human evaluations show that Qwen-Image-2.0 substantially outperforms previous Qwen-Image models in both generation and editing, marking a step toward more general, reliable, and practical image generation foundation models.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Qwen-Image-VAE-2.0 Technical Report

We present Qwen-Image-VAE-2.0, a suite of high-compression Variational Autoencoders (VAEs) that achieve significant advances in both reconstruction fidelity and diffusability. To address the reconstruction bottlenecks of high compression, we adopt an improved architecture featuring Global Skip Connections (GSC) and expanded latent channels. Moreover, we scale training to billions of images and incorporate a synthetic rendering engine to improve performance in text-rich scenarios. To tackle the convergence challenges of high-dimensional latent space, we implement an enhanced semantic alignment strategy to make the latent space highly amenable to diffusion modeling. To optimize computational efficiency, we leverage an asymmetric and attention-free encoder-decoder backbone to minimize encoding overhead. We present a comprehensive evaluation of Qwen-Image-VAE-2.0 on public reconstruction benchmarks. To evaluate performance in text-rich scenarios, we propose OmniDoc-TokenBench, a new benchmark comprising a diverse collection of real-world documents coupled with specialized OCR-based evaluation metrics. Qwen-Image-VAE-2.0 achieves state-of-the-art reconstruction performance, demonstrating exceptional capabilities in both general domains and text-rich scenarios at high compression ratio. Furthermore, downstream DiT experiments reveal our models possess superior diffusability, significantly accelerating convergence compared to existing high-compression baselines. These establish Qwen-Image-VAE-2.0 as a leading model with high compression, superior reconstruction, and exceptional diffusability.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Youtu-LLM: Unlocking the Native Agentic Potential for Lightweight Large Language Models

We introduce Youtu-LLM, a lightweight yet powerful language model that harmonizes high computational efficiency with native agentic intelligence. Unlike typical small models that rely on distillation, Youtu-LLM (1.96B) is pre-trained from scratch to systematically cultivate reasoning and planning capabilities. The key technical advancements are as follows: (1) Compact Architecture with Long-Context Support: Built on a dense Multi-Latent Attention (MLA) architecture with a novel STEM-oriented vocabulary, Youtu-LLM supports a 128k context window. This design enables robust long-context reasoning and state tracking within a minimal memory footprint, making it ideal for long-horizon agent and reasoning tasks. (2) Principled "Commonsense-STEM-Agent" Curriculum: We curated a massive corpus of approximately 11T tokens and implemented a multi-stage training strategy. By progressively shifting the pre-training data distribution from general commonsense to complex STEM and agentic tasks, we ensure the model acquires deep cognitive abilities rather than superficial alignment. (3) Scalable Agentic Mid-training: Specifically for the agentic mid-training, we employ diverse data construction schemes to synthesize rich and varied trajectories across math, coding, and tool-use domains. This high-quality data enables the model to internalize planning and reflection behaviors effectively. Extensive evaluations show that Youtu-LLM sets a new state-of-the-art for sub-2B LLMs. On general benchmarks, it achieves competitive performance against larger models, while on agent-specific tasks, it significantly surpasses existing SOTA baselines, demonstrating that lightweight models can possess strong intrinsic agentic capabilities.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Adapting CLIP For Phrase Localization Without Further Training

Supervised or weakly supervised methods for phrase localization (textual grounding) either rely on human annotations or some other supervised models, e.g., object detectors. Obtaining these annotations is labor-intensive and may be difficult to scale in practice. We propose to leverage recent advances in contrastive language-vision models, CLIP, pre-trained on image and caption pairs collected from the internet. In its original form, CLIP only outputs an image-level embedding without any spatial resolution. We adapt CLIP to generate high-resolution spatial feature maps. Importantly, we can extract feature maps from both ViT and ResNet CLIP model while maintaining the semantic properties of an image embedding. This provides a natural framework for phrase localization. Our method for phrase localization requires no human annotations or additional training. Extensive experiments show that our method outperforms existing no-training methods in zero-shot phrase localization, and in some cases, it even outperforms supervised methods. Code is available at https://github.com/pals-ttic/adapting-CLIP .

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Hybrid Spatial-Temporal Entropy Modelling for Neural Video Compression

For neural video codec, it is critical, yet challenging, to design an efficient entropy model which can accurately predict the probability distribution of the quantized latent representation. However, most existing video codecs directly use the ready-made entropy model from image codec to encode the residual or motion, and do not fully leverage the spatial-temporal characteristics in video. To this end, this paper proposes a powerful entropy model which efficiently captures both spatial and temporal dependencies. In particular, we introduce the latent prior which exploits the correlation among the latent representation to squeeze the temporal redundancy. Meanwhile, the dual spatial prior is proposed to reduce the spatial redundancy in a parallel-friendly manner. In addition, our entropy model is also versatile. Besides estimating the probability distribution, our entropy model also generates the quantization step at spatial-channel-wise. This content-adaptive quantization mechanism not only helps our codec achieve the smooth rate adjustment in single model but also improves the final rate-distortion performance by dynamic bit allocation. Experimental results show that, powered by the proposed entropy model, our neural codec can achieve 18.2% bitrate saving on UVG dataset when compared with H.266 (VTM) using the highest compression ratio configuration. It makes a new milestone in the development of neural video codec. The codes are at https://github.com/microsoft/DCVC.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Neural Compression-Based Feature Learning for Video Restoration

How to efficiently utilize the temporal features is crucial, yet challenging, for video restoration. The temporal features usually contain various noisy and uncorrelated information, and they may interfere with the restoration of the current frame. This paper proposes learning noise-robust feature representations to help video restoration. We are inspired by that the neural codec is a natural denoiser. In neural codec, the noisy and uncorrelated contents which are hard to predict but cost lots of bits are more inclined to be discarded for bitrate saving. Therefore, we design a neural compression module to filter the noise and keep the most useful information in features for video restoration. To achieve robustness to noise, our compression module adopts a spatial channel-wise quantization mechanism to adaptively determine the quantization step size for each position in the latent. Experiments show that our method can significantly boost the performance on video denoising, where we obtain 0.13 dB improvement over BasicVSR++ with only 0.23x FLOPs. Meanwhile, our method also obtains SOTA results on video deraining and dehazing.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Roman: Making Everyday Objects Robotically Manipulable with 3D-Printable Add-on Mechanisms

One important vision of robotics is to provide physical assistance by manipulating different everyday objects, e.g., hand tools, kitchen utensils. However, many objects designed for dexterous hand-control are not easily manipulable by a single robotic arm with a generic parallel gripper. Complementary to existing research on developing grippers and control algorithms, we present Roman, a suite of hardware design and software tool support for robotic engineers to create 3D printable mechanisms attached to everyday handheld objects, making them easier to be manipulated by conventional robotic arms. The Roman hardware comes with a versatile magnetic gripper that can snap on/off handheld objects and drive add-on mechanisms to perform tasks. Roman also provides software support to register and author control programs. To validate our approach, we designed and fabricated Roman mechanisms for 14 everyday objects/tasks presented within a design space and conducted expert interviews with robotic engineers indicating that Roman serves as a practical alternative for enabling robotic manipulation of everyday objects.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Curvy: An Interactive Design Tool for Varying Density Support Structures

We introduce Curvy-an interactive design tool to generate varying density support structures for 3D printing. Support structures are essential for printing models with extreme overhangs. Yet, they often cause defects on contact areas, resulting in poor surface quality. Low-level design of support structures may alleviate such negative effects. However, it is tedious and unintuitive for novice users as it is hard to predict the impact of changes to the support structure on the final printed part. Curvy allows users to define their high-level preferences on the surface quality directly on the target object rather than explicitly designing the supports. These preferences are then automatically translated into low-level design parameters to generate the support structure. Underlying novel curvy zigzag toolpathing algorithm uses these instructions to generate varying density supports by altering the spacing between individual paths in order to achieve prescribed quality. Combined with the build orientation optimization, Curvy provides a practical solution to the design of support structures with minimal perceptual or functional impact on the target part to be printed.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Adaptive Generation Model: A New Ensemble Method

As a common method in Machine Learning, Ensemble Method is used to train multiple models from a data set and obtain better results through certain combination strategies. Stacking method, as representatives of Ensemble Learning methods, is often used in Machine Learning Competitions such as Kaggle. This paper proposes a variant of Stacking Model based on the idea of gcForest, namely Adaptive Generation Model (AGM). It means that the adaptive generation is performed not only in the horizontal direction to expand the width of each layer model, but also in the vertical direction to expand the depth of the model. For base models of AGM, they all come from preset basic Machine Learning Models. In addition, a feature augmentation method is added between layers to further improve the overall accuracy of the model. Finally, through comparative experiments on 7 data sets, the results show that the accuracy of AGM are better than its previous models.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Iterative Distance-Aware Similarity Matrix Convolution with Mutual-Supervised Point Elimination for Efficient Point Cloud Registration

In this paper, we propose a novel learning-based pipeline for partially overlapping 3D point cloud registration. The proposed model includes an iterative distance-aware similarity matrix convolution module to incorporate information from both the feature and Euclidean space into the pairwise point matching process. These convolution layers learn to match points based on joint information of the entire geometric features and Euclidean offset for each point pair, overcoming the disadvantage of matching by simply taking the inner product of feature vectors. Furthermore, a two-stage learnable point elimination technique is presented to improve computational efficiency and reduce false positive correspondence pairs. A novel mutual-supervision loss is proposed to train the model without extra annotations of keypoints. The pipeline can be easily integrated with both traditional (e.g. FPFH) and learning-based features. Experiments on partially overlapping and noisy point cloud registration show that our method outperforms the current state-of-the-art, while being more computationally efficient. Code is publicly available at https://github.com/jiahaowork/idam.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Romeo: A Design Tool for Embedding Transformable Parts in 3D Models to Robotically Augment Default Functionalities

Reconfiguring shapes of objects enables transforming existing passive objects with robotic functionalities, e.g., a transformable coffee cup holder can be attached to a chair's armrest, a piggy bank can reach out an arm to 'steal' coins. Despite the advance in end-user 3D design and fabrication, it remains challenging for non-experts to create such 'transformables' using existing tools due to the requirement of specific engineering knowledge such as mechanisms and robotic design. We present Romeo -- a design tool for creating transformables to robotically augment objects' default functionalities. Romeo allows users to transform an object into a robotic arm by expressing at a high level what type of task is expected. Users can select which part of the object to be transformed, specify motion points in space for the transformed part to follow and the corresponding action to be taken. Romeo then automatically generates a robotic arm embedded in the transformable part ready for fabrication. A design session validated this tool where participants used Romeo to accomplish controlled design tasks and to open-endedly create coin-stealing piggy banks by transforming 3D objects of their own choice.

0 citations0 reviewsNo code linkedOpen access
preprint2018arXiv

Network architecture for a topological quantum computer in silicon

A design for a large-scale surface code quantum processor based on a node/network approach is introduced for semiconductor quantum dot spin qubits. The minimal node contains only 7 quantum dots, and nodes are separated on the micron scale, creating useful space for wiring interconnects and integration of conventional transistor circuits. Entanglement is distributed between neighbouring nodes by loading spin singlets locally and then shuttling one member of the pair through a linear array of empty dots. Each node contains one data qubit, two ancilla qubits, and additional dots to facilitate electron shuttling and measurement of the ancillas. A four-node GHZ state is realized by sharing three internode singlets followed by local gate operations and ancilla measurements. Further local operations and measurements produce an X or Z stabilizer on four data qubits, which is the fundamental operation of the surface code. Electron shuttling is simulated using a simplified gate electrode geometry without explicit barrier gates, and demonstrates that adiabatic transport is possible on timescales that do not present a speed bottleneck to the processor. An important shuttling error in a clean system is uncontrolled phase rotation due to the modulation of the electronic g-factor during transport, owing to the Stark effect. This error can be reduced by appropriate electrostatic tuning of the stationary electron's g-factor. Using reasonable noise models, we estimate error thresholds with respect to single and two-qubit gate fidelities as well as singlet dephasing errors during shuttling. A twirling protocol transforms the non-Pauli noise associated with exchange gate operations into Pauli noise, making it possible to use the Gottesman-Knill theorem to efficiently simulate large codes.

0 citations0 reviewsNo code linkedOpen access