arXiv subjects

Haibo Qiu

Publications and source records attributed to Haibo Qiu.

At least 37 records · Page 2Linked to original sources

arXiv2025

Perceptual-Evidence Anchored Reinforced Learning for Multimodal Reasoning

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Large Language Models (LLMs) and is now being applied to Vision-Language Models (VLMs). However, vanilla RLVR for VLMs verifies only the final textual output, critically neglecting the foundational step of visual perception. This oversight leads to visual hallucinations and reward hacking, as reasoning built upon flawed perception is inherently unreliable. To address this, we propose PEARL (Perceptual-Evidence Anchored Reinforced Learning), a dual-branch, perception-reasoning synergistic that strengthens multimodal reasoning by explicitly anchoring it to verified visual evidence. For each reasoning-oriented QA instance, PEARL first derive a perception checklist -- a set of perception-oriented sub-questions with verifiable answers that probe the model's understanding of key visual evidence. During training, auxiliary rollouts on this checklist yield a perceptual reward that both directly reinforces the model's perception ability and acts as a fidelity gate for reasoning. If the model passes the perception check, its policy update is biased towards evidence-anchored reasoning. Otherwise, the process is halted to prevent reasoning from flawed premises. PEARL can be seamlessly integrated with popular RL methods like GRPO and DAPO. Comprehensive experiments show PEARL achieves substantial gains on multimodal reasoning benchmarks, e.g., a +9.7% improvement over the baseline and +6.6% over GRPO on MathVerse.

cs.CV↗

arXiv2025

Towards Better & Faster Autoregressive Image Generation: From the Perspective of Entropy

In this work, we first revisit the sampling issues in current autoregressive (AR) image generation models and identify that image tokens, unlike text tokens, exhibit lower information density and non-uniform spatial distribution. Accordingly, we present an entropy-informed decoding strategy that facilitates higher autoregressive generation quality with faster synthesis speed. Specifically, the proposed method introduces two main innovations: 1) dynamic temperature control guided by spatial entropy of token distributions, enhancing the balance between content diversity, alignment accuracy, and structural coherence in both mask-based and scale-wise models, without extra computational overhead, and 2) entropy-aware acceptance rules in speculative decoding, achieving near-lossless generation at about 85\% of the inference cost of conventional acceleration methods. Extensive experiments across multiple benchmarks using diverse AR image generation models demonstrate the effectiveness and generalizability of our approach in enhancing both generation quality and sampling speed.

cs.CV↗

arXiv2025

STAGE: Stable and Generalizable GRPO for Autoregressive Image Generation

Reinforcement learning has recently been explored to improve text-to-image generation, yet applying existing GRPO algorithms to autoregressive (AR) image models remains challenging. The instability of the training process easily disrupts the pretrained model capability during long runs, resulting in marginal gains, degraded image quality, and poor generalization. In this work, we revisit GRPO for AR image generation and identify two key issues: contradictory gradients from unnecessary tokens and unstable policy entropy dynamics. To address these, we introduce STAGE, a stable and generalizable framework that leverages two targeted solutions: 1) Advantage/KL reweighting. Similarity-aware reweighting to alleviate conflicting updates; and 2) Entropy reward. An entropy-based reward corresponding to reference model to stabilize learning. With the help of alleviating conflicts between tokens and an entropy reward for stabilizing training, we reduce disruption of the pretrained distribution and mitigate reward hacking, which in turn improves generalization and transfer better to other benchmarks. Experiments across multiple benchmarks show that STAGE consistently improves visual quality, stability, and cross-task generalization compared to baseline GRPO.

cs.CV↗

arXiv2025

OmniActor: A Generalist GUI and Embodied Agent for 2D&3D Worlds

Multimodal large language models are evolving toward multimodal agents capable of proactively executing tasks. Most agent research focuses on GUI or embodied scenarios, which correspond to agents interacting with 2D virtual worlds or 3D real worlds, respectively. However, many complex tasks typically require agents to interleavely interact with these two types of environment. We initially mix GUI and embodied data to train, but find the performance degeneration brought by the data conflict. Further analysis reveals that GUI and embodied data exhibit synergy and conflict at the shallow and deep layers, respectively, which resembles the cerebrum-cerebellum mechanism in the human brain. To this end, we propose a high-performance generalist agent OmniActor, designed from both structural and data perspectives. First, we propose Layer-heterogeneity MoE to eliminate the conflict between GUI and embodied data by separating deep-layer parameters, while leverage their synergy by sharing shallow-layer parameters. By successfully leveraging the synergy and eliminating the conflict, OmniActor outperforms agents only trained by GUI or embodied data in GUI or embodied tasks. Furthermore, we unify the action spaces of GUI and embodied tasks, and collect large-scale GUI and embodied data from various sources for training. This significantly improves OmniActor under different scenarios, especially in GUI tasks. The code will be publicly available.

cs.CV↗

arXiv2025

Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning

Recent advancements in large language models (LLMs) have witnessed a surge in the development of advanced reasoning paradigms, which are now being integrated into multimodal large language models (MLLMs). However, existing approaches often fall short: methods solely employing reinforcement learning (RL) can struggle with sample inefficiency and activating entirely absent reasoning capabilities, while conventional pipelines that initiate with a cold-start supervised fine-tuning (SFT) phase before RL may restrict the model's exploratory capacity and face suboptimal convergence. In this work, we introduce \textbf{Metis-RISE} (\textbf{R}L \textbf{I}ncentivizes and \textbf{S}FT \textbf{E}nhances) for multimodal reasoning model learning. Unlike conventional approaches, Metis-RISE distinctively omits an initial SFT stage, beginning instead with an RL phase (e.g., using a Group Relative Policy Optimization variant) to incentivize and activate the model's latent reasoning capacity. Subsequently, the targeted SFT stage addresses two key challenges identified during RL: (1) \textit{inefficient trajectory sampling} for tasks where the model possesses but inconsistently applies correct reasoning, which we tackle using self-distilled reasoning trajectories from the RL model itself; and (2) \textit{fundamental capability absence}, which we address by injecting expert-augmented knowledge for prompts where the model entirely fails. This strategic application of RL for incentivization followed by SFT for enhancement forms the core of Metis-RISE, leading to two versions of our MLLMs (7B and 72B parameters). Evaluations on the OpenCompass Multimodal Reasoning Leaderboard demonstrate that both models achieve state-of-the-art performance among similar-sized models, with the 72B version ranking fourth overall. Please refer to our project page for open-source information.

cs.AI↗

arXiv2025

UniToken: Harmonizing Multimodal Understanding and Generation through Unified Visual Encoding

We introduce UniToken, an auto-regressive generation model that encodes visual inputs through a combination of discrete and continuous representations, enabling seamless integration of unified visual understanding and image generation tasks. Unlike previous approaches that rely on unilateral visual representations, our unified visual encoding framework captures both high-level semantics and low-level details, delivering multidimensional information that empowers heterogeneous tasks to selectively assimilate domain-specific knowledge based on their inherent characteristics. Through in-depth experiments, we uncover key principles for developing a unified model capable of both visual understanding and image generation. Extensive evaluations across a diverse range of prominent benchmarks demonstrate that UniToken achieves state-of-the-art performance, surpassing existing approaches. These results establish UniToken as a robust foundation for future research in this domain. The code and models are available at https://github.com/SxJyJay/UniToken.

cs.CV↗

arXiv2024

Hysteresis loops in spinor BECs subject to synthetic gauge fields

We explore the hysteretic dynamics of spinor Bose-Einstein condensates of ultracold atoms loaded in static 2D ring geometries and subjected to varying synthetic magnetic fields. Electrically neutral, pseudo-spin-$1/2$ condensates are probed by one-component weak-link potentials that make available paths for the transit of vortices into and out of the ring in both components, and thus can control the number of flux quanta threading the ring hole. For perpendicular fields, domain walls that are dynamically generated in the relative phase of the spin components are shown to play a key role in driving the hysteretic behaviour of the whole system through changes in the net magnetic flux. In the presence of spin-orbit-coupling, hysteresis is exhibited around the phase transitions found by the metastable current states for varying fields.

cond-mat.quant-gas↗

arXiv2024

Scalable synchronization cluster in networked chaotic oscillators

Cluster synchronization in synthetic networks of coupled chaotic oscillators is investigated. It is found that despite the asymmetric nature of the network structure, a subset of the oscillators can be synchronized as a cluster while the other oscillators remain desynchronized. Interestingly, with the increase of the coupling strength, the cluster is expanding gradually by recruiting the desynchronized oscillators one by one. This new synchronization phenomenon, which is named ``scalable synchronization cluster", is explored theoretically by the method of eigenvector-based analysis, and it is revealed that the scalability of the cluster is attributed to the unique feature of the eigenvectors of the network coupling matrix. The transient dynamics of the cluster in response to random perturbations are also studied, and it is shown that in restoring to the synchronization state, oscillators inside the cluster are stabilized in sequence, illustrating again the hierarchy of the oscillators. The findings shed new light on the collective behaviors of networked chaotic oscillators, and are helpful for the design of real-world networks where scalable synchronization clusters are concerned.

nlin.AO↗

arXiv2023

Collect-and-Distribute Transformer for 3D Point Cloud Analysis

Remarkable advancements have been made recently in point cloud analysis through the exploration of transformer architecture, but it remains challenging to effectively learn local and global structures within point clouds. In this paper, we propose a new transformer network equipped with a collect-and-distribute mechanism to communicate short- and long-range contexts of point clouds, which we refer to as CDFormer. Specifically, we first employ self-attention to capture short-range interactions within each local patch, and the updated local features are then collected into a set of proxy reference points from which we can extract long-range contexts. Afterward, we distribute the learned long-range contexts back to local points via cross-attention. To address the position clues for short- and long-range contexts, we additionally introduce the context-aware position encoding to facilitate position-aware communications between points. We perform experiments on five popular point cloud datasets, namely ModelNet40, ScanObjectNN, ShapeNetPart, S3DIS and ScanNetV2, for classification and segmentation. Results show the effectiveness of the proposed CDFormer, delivering several new state-of-the-art performances on point cloud classification and segmentation tasks. The source code is available at \url{https://github.com/haibo-qiu/CDFormer}.

cs.CV↗

arXiv2023

Gap solitons and nonlinear Bloch states in Bose-Einstein condensates with current-dependent interactions

We show how the chiral properties of Bose Einstein condensates subject to current-density interactions and loaded in optical lattices can be observed in the realization of nonlinear Bloch states, whose spectrum lacks the usual periodic structure. Chirality is also manifested by spatially localized states, or gap solitons, which are found for positive rotation rates of the lattice at the energy gaps between the linear energy bands, whereas for negative rotations they appear in the semi-infinite gap of the linear spectrum. The stability of extended and localized states is checked through the spectrum of linear excitations and nonlinear time evolution of perturbed states, and the phenomenon of Bloch oscillations is explored. Our results are obtained in quasi 1D ring geometries with feasible experimental parameters.

cond-mat.quant-gas↗

arXiv2023

PointHR: Exploring High-Resolution Architectures for 3D Point Cloud Segmentation

Significant progress has been made recently in point cloud segmentation utilizing an encoder-decoder framework, which initially encodes point clouds into low-resolution representations and subsequently decodes high-resolution predictions. Inspired by the success of high-resolution architectures in image dense prediction, which always maintains a high-resolution representation throughout the entire learning process, we consider it also highly important for 3D dense point cloud analysis. Therefore, in this paper, we explore high-resolution architectures for 3D point cloud segmentation. Specifically, we generalize high-resolution architectures using a unified pipeline named PointHR, which includes a knn-based sequence operator for feature extraction and a differential resampling operator to efficiently communicate different resolutions. Additionally, we propose to avoid numerous on-the-fly computations of high-resolution architectures by pre-computing the indices for both sequence and resampling operators. By doing so, we deliver highly competitive high-resolution architectures while capitalizing on the benefits of well-designed point cloud blocks without additional effort. To evaluate these architectures for dense point cloud analysis, we conduct thorough experiments using S3DIS and ScanNetV2 datasets, where the proposed PointHR outperforms recent state-of-the-art methods without any bells and whistles. The source code is available at \url{https://github.com/haibo-qiu/PointHR}.

cs.CV↗

arXiv2022

GFNet: Geometric Flow Network for 3D Point Cloud Semantic Segmentation

Point cloud semantic segmentation from projected views, such as range-view (RV) and bird's-eye-view (BEV), has been intensively investigated. Different views capture different information of point clouds and thus are complementary to each other. However, recent projection-based methods for point cloud semantic segmentation usually utilize a vanilla late fusion strategy for the predictions of different views, failing to explore the complementary information from a geometric perspective during the representation learning. In this paper, we introduce a geometric flow network (GFNet) to explore the geometric correspondence between different views in an align-before-fuse manner. Specifically, we devise a novel geometric flow module (GFM) to bidirectionally align and propagate the complementary information across different views according to geometric relationships under the end-to-end learning scheme. We perform extensive experiments on two widely used benchmark datasets, SemanticKITTI and nuScenes, to demonstrate the effectiveness of our GFNet for project-based point cloud semantic segmentation. Concretely, GFNet not only significantly boosts the performance of each individual view but also achieves state-of-the-art results over all existing projection-based models. Code is available at \url{https://github.com/haibo-qiu/GFNet}.

cs.CV↗

arXiv2022

Soliton collisions in Bose-Einstein condensates with current-dependent interactions

We study general collisions between chiral solitons in Bose-Einstein condensates subject to combined attractive and current-dependent interatomic interactions. A simple analysis based on the linear superposition of the solitons allows us to determine the relevant time and space scales of the dynamics, which is illustrated by extensive numerical simulations. By varying the differential amplitude, the relative phase, the average velocity, and the relative velocity of the solitons, we characterize the different dynamical regimes that give rise to oscillatory and interference phenomena. Apart from the known inelastic character of the collisions, we show that the chiral dynamics involves an amplitude reduction with respect to the case of regular solitons. To compare with feasible ultracold gas experiments, the influence of harmonic confinement is analyzed in both the emergence and the interaction of chiral solitons.

cond-mat.quant-gas↗

arXiv2022

End2End Occluded Face Recognition by Masking Corrupted Features

With the recent advancement of deep convolutional neural networks, significant progress has been made in general face recognition. However, the state-of-the-art general face recognition models do not generalize well to occluded face images, which are exactly the common cases in real-world scenarios. The potential reasons are the absences of large-scale occluded face data for training and specific designs for tackling corrupted features brought by occlusions. This paper presents a novel face recognition method that is robust to occlusions based on a single end-to-end deep neural network. Our approach, named FROM (Face Recognition with Occlusion Masks), learns to discover the corrupted features from the deep convolutional neural networks, and clean them by the dynamically learned masks. In addition, we construct massive occluded face images to train FROM effectively and efficiently. FROM is simple yet powerful compared to the existing methods that either rely on external detectors to discover the occlusions or employ shallow models which are less discriminative. Experimental results on the LFW, Megaface challenge 1, RMF2, AR dataset and other simulated occluded/masked datasets confirm that FROM dramatically improves the accuracy under occlusions, and generalizes well on general face recognition. Code is available at https://github.com/haibo-qiu/FROM

cs.CV↗

arXiv2021

SynFace: Face Recognition with Synthetic Data

With the recent success of deep neural networks, remarkable progress has been achieved on face recognition. However, collecting large-scale real-world training data for face recognition has turned out to be challenging, especially due to the label noise and privacy issues. Meanwhile, existing face recognition datasets are usually collected from web images, lacking detailed annotations on attributes (e.g., pose and expression), so the influences of different attributes on face recognition have been poorly investigated. In this paper, we address the above-mentioned issues in face recognition using synthetic face images, i.e., SynFace. Specifically, we first explore the performance gap between recent state-of-the-art face recognition models trained with synthetic and real face images. We then analyze the underlying causes behind the performance gap, e.g., the poor intra-class variations and the domain gap between synthetic and real face images. Inspired by this, we devise the SynFace with identity mixup (IM) and domain mixup (DM) to mitigate the above performance gap, demonstrating the great potentials of synthetic data for face recognition. Furthermore, with the controllable face synthesis model, we can easily manage different factors of synthetic face generation, including pose, expression, illumination, the number of identities, and samples per identity. Therefore, we also perform a systematically empirical analysis on synthetic face images to provide some insights on how to effectively utilize synthetic data for face recognition.

cs.CV↗

arXiv2021

Stripe and junction-vortex phases in linearly coupled Bose-Einstein condensates

Soon after its theoretical prediction, striped-density states in the presence of synthetic spin-orbit coupling were realized in Bose-Einstein condensates of ultracold neutral atoms [J.-R. Li et al., Nature \textbf{543}, 91 (2017)]. The achievement opens avenues to explore the interplay of superfluidity and crystalline order in the search for supersolid features and materials. The system considered is essentially made of two linearly coupled Bose-Einstein condensates, that is a pseudo-spin-$1/2$ system, subject to a spin-dependent gauge field $σ_z \hbar k_\ell$. Under these conditions the stripe phase is achieved when the linear coupling $\hbarΩ/2$ is small against the gauge energy $mΩ/\hbar k_\ell^2<1$ . The resulting density stripes have been interpreted as a standing-wave, interference pattern with approximate wavenumber $2k_\ell$. Here, we show that the emergence of the stripe phase is induced by an array of Josephson vortices living in the junction defined by the linear coupling. As happens in superconducting junctions subject to external magnetic fields, a vortex array is the natural response of the superfluid system to the presence of a gauge field. Also similar to superconductors, the Josephson currents and their associated vortices can be present as a metastable state in the absence of gauge field. We provide closed-form solutions to the 1D mean field equations that account for such vortex arrays. The underlying Josephson currents coincide with the analytical solutions to the sine-Gordon equation for the relative phase of superconducting junctions [C. Owen and D. Scalapino, Phys. Rev. \textbf{164}, 538 (1967)].

cond-mat.quant-gas↗

arXiv2020

Unlocked-relative-phase states in arrays of Bose-Einstein condensates

Phase engineering techniques are used to control the dynamics of long-bosonic-Josephson-junction arrays built by linearly coupling Bose-Einstein condensates. Just at the middle point of the underlying discrete energy band of the system, unlocked-relative-phase states are shown to be stationary along with the locked-relative-phase Bloch waves. In finite, experimentally-feasible systems, such states find ranges of dynamical stability that depend on the ratio of coupling to interaction energy. The same ratio determines different decay regimes, which include the recurrence of staggered-soliton trains in the condensates around Josephson loop currents at the junctions. These transient solitons are also found in their stationary configurations, which provide striped-density states by means of either dark-soliton or bright-soliton trains. Additionally, the preparation of maximally out-of-phase (or splay) states is demonstrated to evolve into an oscillation of the uniform density of the condensates that keeps constant the total density of the system and robust against noise at low coupling.

cond-mat.quant-gas↗

arXiv2019

Cross View Fusion for 3D Human Pose Estimation

We present an approach to recover absolute 3D human poses from multi-view images by incorporating multi-view geometric priors in our model. It consists of two separate steps: (1) estimating the 2D poses in multi-view images and (2) recovering the 3D poses from the multi-view 2D poses. First, we introduce a cross-view fusion scheme into CNN to jointly estimate 2D poses for multiple views. Consequently, the 2D pose estimation for each view already benefits from other views. Second, we present a recursive Pictorial Structure Model to recover the 3D pose from the multi-view 2D poses. It gradually improves the accuracy of 3D pose with affordable computational cost. We test our method on two public datasets H36M and Total Capture. The Mean Per Joint Position Errors on the two datasets are 26mm and 29mm, which outperforms the state-of-the-arts remarkably (26mm vs 52mm, 29mm vs 35mm). Our code is released at \url{https://github.com/microsoft/multiview-human-pose-estimation-pytorch}.

cs.CV↗