Searcharxiv⌕ Search

arXiv subjects

Yi Yang

Publications and source records attributed to Yi Yang.

At least 505 records · Page 28Linked to original sources
arXiv2024

FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention

Video diffusion models have made substantial progress in various video generation applications. However, training models for long video generation tasks require significant computational and data resources, posing a challenge to developing long video diffusion models. This paper investigates a straightforward and training-free approach to extend an existing short video diffusion model (e.g. pre-trained on 16-frame videos) for consistent long video generation (e.g. 128 frames). Our preliminary observation has found that directly applying the short video diffusion model to generate long videos can lead to severe video quality degradation. Further investigation reveals that this degradation is primarily due to the distortion of high-frequency components in long videos, characterized by a decrease in spatial high-frequency components and an increase in temporal high-frequency components. Motivated by this, we propose a novel solution named FreeLong to balance the frequency distribution of long video features during the denoising process. FreeLong blends the low-frequency components of global video features, which encapsulate the entire video sequence, with the high-frequency components of local video features that focus on shorter subsequences of frames. This approach maintains global consistency while incorporating diverse and high-quality spatiotemporal details from local videos, enhancing both the consistency and fidelity of long video generation. We evaluated FreeLong on multiple base video diffusion models and observed significant improvements. Additionally, our method supports coherent multi-prompt generation, ensuring both visual coherence and seamless transitions between scenes.

cs.CV↗
arXiv2024

PiPa++: Towards Unification of Domain Adaptive Semantic Segmentation via Self-supervised Learning

Unsupervised domain adaptive segmentation aims to improve the segmentation accuracy of models on target domains without relying on labeled data from those domains. This approach is crucial when labeled target domain data is scarce or unavailable. It seeks to align the feature representations of the source domain (where labeled data is available) and the target domain (where only unlabeled data is present), thus enabling the model to generalize well to the target domain. Current image- and video-level domain adaptation have been addressed using different and specialized frameworks, training strategies and optimizations despite their underlying connections. In this paper, we propose a unified framework PiPa++, which leverages the core idea of ``comparing'' to (1) explicitly encourage learning of discriminative pixel-wise features with intraclass compactness and inter-class separability, (2) promote the robust feature learning of the identical patch against different contexts or fluctuations, and (3) enable the learning of temporal continuity under dynamic environments. With the designed task-smart contrastive sampling strategy, PiPa++ enables the mining of more informative training samples according to the task demand. Extensive experiments demonstrate the effectiveness of our method on both image-level and video-level domain adaption benchmarks. Moreover, the proposed method is compatible with other UDA approaches to further improve the performance without introducing extra parameters.

cs.CV↗
arXiv2024

Probing black holes in a dark matter spike of M87 using quasinormal modes

Dark matter density can be significantly enhanced by the supermassive black hole at the galactic center, leading to a structure called dark matter spike. Dark matter spike may change the spacetime properties of black holes that constitute deviations from GR black holes. Based on these interesting background, we construct a set of solutions of black holes in a dark matter spike under the Newtonian approximation and full relativity. Combining the mass model of M87, we study the quasinormal modes of black holes in the scalar field and axial gravitational perturbation, then compared them with Schwarzschild black hole. Besides, the impacts of dark matter on the quasinormal mode of black holes have been studied in depth. In particular, in the axial gravitational perturbation, our results show that the impacts of dark matter spike on the quasinormal mode of black holes can reach up to $10^{-4}$. These new features from quasinormal mode of black holes under the Newtonian approximation and full relativity may provide some help for the establishment of the final dark matter model, and provide a new thought for the indirect detection of dark matter.

gr-qc↗
arXiv2024

Early-Time Observations of SN 2023wrk: A Luminous Type Ia Supernova with Significant Unburned Carbon in the Outer Ejecta

We present extensive photometric and spectroscopic observations of the nearby Type Ia supernova (SN) 2023wrk at a distance of about 40 Mpc. The earliest detection of this SN can be traced back to a few hours after the explosion. Within the first few days the light curve shows a bump feature, while the B - V color is blue and remains nearly constant. The overall spectral evolution is similar to that of an SN 1991T/SN 1999aa-like SN Ia, while the C II $\lambda6580$ absorption line appears to be unusually strong in the first spectrum taken at $t \approx -$15.4 days after the maximum light. This carbon feature disappears quickly in subsequent evolution but it reappears at around the time of peak brightness. The complex evolution of the carbon line and the possible detection of Ni III absorption around 4700 Å and 5300 Å in the earliest spectra indicate macroscopic mixing of fuel and ash. The strong carbon lines is likely related to collision of SN ejecta with unbound carbon, consistent with the predictions of pulsational delayed-detonation or carbon-rich circumstellar-matter interaction models. Among those carbon-rich SNe Ia with strong C II $\lambda6580$ absorption at very early times, the line-strength ratio of C II to Si II and the B-V color evolution are found to exhibit large diversity, which may be attributed to different properties of unbound carbon and outward-mixing $^{56}$Ni.

astro-ph.SR↗
arXiv2024

Navigation Instruction Generation with BEV Perception and Large Language Models

Navigation instruction generation, which requires embodied agents to describe the navigation routes, has been of great interest in robotics and human-computer interaction. Existing studies directly map the sequence of 2D perspective observations to route descriptions. Though straightforward, they overlook the geometric information and object semantics of the 3D environment. To address these challenges, we propose BEVInstructor, which incorporates Bird's Eye View (BEV) features into Multi-Modal Large Language Models (MLLMs) for instruction generation. Specifically, BEVInstructor constructs a PerspectiveBEVVisual Encoder for the comprehension of 3D environments through fusing BEV and perspective features. To leverage the powerful language capabilities of MLLMs, the fused representations are used as visual prompts for MLLMs, and perspective-BEV prompt tuning is proposed for parameter-efficient updating. Based on the perspective-BEV prompts, BEVInstructor further adopts an instance-guided iterative refinement pipeline, which improves the instructions in a progressive manner. BEVInstructor achieves impressive performance across diverse datasets (i.e., R2R, REVERIE, and UrbanWalk).

cs.CV↗
arXiv2024

Probing the Shock Breakout Signal of SN 2024ggi from the Transformation of Early Flash Spectroscopy

We present early-time, hour-to-day cadence spectroscopy of the nearby type II supernova (SN II) 2024ggi, which was discovered at a phase when the SN shock just emerged from the red-supergiant (RSG) progenitor star. Over the first few days after the first light, SN 2024ggi exhibited prominent narrow emission lines formed through intense and persistent photoionization of the nearby circumstellar material (CSM). In the first 63 hours, spectral lines of He, C, N, and O revealed a rapid rise in ionization, as a result of the progressive sweeping-up of the CSM by the shock. The duration of the IIn-like spectra indicates a dense and relatively confined CSM distribution extending up to $\sim 4 \times 10^{14}$ cm. Spectral modeling reveals a CSM mass loss rate at this region exceeding $5 \times 10^{-3}{\rm M}_{\odot}$ yr$^{-1}$ is required to reproduce low-ionization emissions, which dramatically exceeds that of an RSG. Analyzing H$α$ emission shift implies the velocity of the unshocked outer CSM to be between 20 and 40 km s$^{-1}$, matching the typical wind velocity of an RSG. The differences between the inner and outer layers of the CSM and an RSG progenitor highlight a complex mass loss history before the explosion of SN 2024ggi.

astro-ph.HE↗
arXiv2024

The Long-lived Broadband Afterglow of Short Gamma-Ray Burst 231117A and the Growing Radio-Detected Short GRB Population

We present multiwavelength observations of the Swift short $γ$-ray burst GRB 231117A, localized to an underlying galaxy at redshift $z = 0.257$ at a small projected offset ($\sim 2~$kpc). We uncover long-lived X-ray (Chandra) and radio/millimeter (VLA, MeerKAT, and ALMA) afterglow emission, detected to $\sim 37~$days and $\sim 20~$days (rest frame), respectively. We measure a wide jet ($\sim 10.4^\circ$) and relatively high circumburst density ($\sim 0.07~{\rm cm}^{-3}$) compared to the short GRB population. Our data cannot be easily fit with a standard forward shock model, but they are generally well fit with the incorporation of a refreshed forward shock and a reverse shock at $< 1~$day. We incorporate GRB 231117A into a larger sample of 132 X-ray detected events, 71 of which were radio-observed (17 cm-band detections), for a systematic study of the distributions of redshifts, jet and afterglow properties, galactocentric offsets, and local environments of events with and without detected radio afterglows. Compared to the entire short GRB population, the majority of radio-detected GRBs are at relatively low redshifts ($z < 0.6$) and have high circumburst densities ($> 10^{-2}~{\rm cm}^{-3}$), consistent with their smaller ($< 8~$kpc) projected galactocentric offsets. We additionally find that 70% of short GRBs with opening angle measurements were radio-detected, indicating the importance of radio afterglows in jet measurements, especially in the cases of wide ($> 10^\circ$) jets where observational evidence of collimation may only be detectable at radio wavelengths. Owing to improved observing strategies and the emergence of sensitive radio facilities, the number of radio-detected short GRBs has quadrupled in the past decade.

astro-ph.HE↗
arXiv2024

VividDreamer: Invariant Score Distillation For Hyper-Realistic Text-to-3D Generation

This paper presents Invariant Score Distillation (ISD), a novel method for high-fidelity text-to-3D generation. ISD aims to tackle the over-saturation and over-smoothing problems in Score Distillation Sampling (SDS). In this paper, SDS is decoupled into a weighted sum of two components: the reconstruction term and the classifier-free guidance term. We experimentally found that over-saturation stems from the large classifier-free guidance scale and over-smoothing comes from the reconstruction term. To overcome these problems, ISD utilizes an invariant score term derived from DDIM sampling to replace the reconstruction term in SDS. This operation allows the utilization of a medium classifier-free guidance scale and mitigates the reconstruction-related errors, thus preventing the over-smoothing and over-saturation of results. Extensive experiments demonstrate that our method greatly enhances SDS and produces realistic 3D objects through single-stage optimization.

cs.CV↗
arXiv2024

SN 2021dbg: A Luminous Type IIP-IIL Supernova Exploding from a Massive Star with a Layered Shell

We present extensive observations and analysis of supernova (SN) 2021dbg, utilizing optical photometry and spectroscopy. For approximately 385 days following the explosion, SN 2021dbg exhibited remarkable luminosity, surpassing most SNe II. This initial high luminosity is potentially attributed to the interaction between the ejected material and the surrounding circumstellar material (CSM), as evidenced by the pronounced interaction signatures observed in its spectra. The subsequent high luminosity is primarily due to the significant $^{56}$Ni ($0.17 \pm 0.05$ M$_{\odot}$) produced in the explosion. Based on the flux of flash emission lines detected in the initial spectra, we estimate that the CSM mass near the progenitor amounted to $\sim$(1.0--2.0) $\times 10^{-3}$ M$_{\odot}$, likely resulting from intense stellar wind activity 2--3 yr preceding the explosion. Considering the bolometric light curve, nebular spectrum modeling, and mass-loss rate, we suggest that the progenitor of SN 2021dbg was a red supergiant (RSG) with a mass of $\sim 20$ M$_{\odot}$ and a radius of 1200 R$_{\odot}$. This RSG featured a thick hydrogen shell, which may have contained a region with a sharp decrease in material density, electron density, and temperature, contributing to its layered structure. This object demonstrates mixed features of SNe IIP and SNe IIL, making it as a transitional event linking the above two subclasses of SNe II.

astro-ph.HE↗
arXiv2024

Controllable Navigation Instruction Generation with Chain of Thought Prompting

Instruction generation is a vital and multidisciplinary research area with broad applications. Existing instruction generation models are limited to generating instructions in a single style from a particular dataset, and the style and content of generated instructions cannot be controlled. Moreover, most existing instruction generation methods also disregard the spatial modeling of the navigation environment. Leveraging the capabilities of Large Language Models (LLMs), we propose C-Instructor, which utilizes the chain-of-thought-style prompt for style-controllable and content-controllable instruction generation. Firstly, we propose a Chain of Thought with Landmarks (CoTL) mechanism, which guides the LLM to identify key landmarks and then generate complete instructions. CoTL renders generated instructions more accessible to follow and offers greater controllability over the manipulation of landmark objects. Furthermore, we present a Spatial Topology Modeling Task to facilitate the understanding of the spatial structure of the environment. Finally, we introduce a Style-Mixed Training policy, harnessing the prior knowledge of LLMs to enable style control for instruction generation based on different prompts within a single model instance. Extensive experiments demonstrate that instructions generated by C-Instructor outperform those generated by previous methods in text metrics, navigation guidance evaluation, and user studies.

cs.CV↗
arXiv2024

Mutual Learning for Acoustic Matching and Dereverberation via Visual Scene-driven Diffusion

Visual acoustic matching (VAM) is pivotal for enhancing the immersive experience, and the task of dereverberation is effective in improving audio intelligibility. Existing methods treat each task independently, overlooking the inherent reciprocity between them. Moreover, these methods depend on paired training data, which is challenging to acquire, impeding the utilization of extensive unpaired data. In this paper, we introduce MVSD, a mutual learning framework based on diffusion models. MVSD considers the two tasks symmetrically, exploiting the reciprocal relationship to facilitate learning from inverse tasks and overcome data scarcity. Furthermore, we employ the diffusion model as foundational conditional converters to circumvent the training instability and over-smoothing drawbacks of conventional GAN architectures. Specifically, MVSD employs two converters: one for VAM called reverberator and one for dereverberation called dereverberator. The dereverberator judges whether the reverberation audio generated by reverberator sounds like being in the conditional visual scenario, and vice versa. By forming a closed loop, these two converters can generate informative feedback signals to optimize the inverse tasks, even with easily acquired one-way unpaired data. Extensive experiments on two standard benchmarks, i.e., SoundSpaces-Speech and Acoustic AVSpeech, exhibit that our framework can improve the performance of the reverberator and dereverberator and better match specified visual scenarios.

cs.SD↗
arXiv2024

Nonverbal Interaction Detection

This work addresses a new challenge of understanding human nonverbal interaction in social contexts. Nonverbal signals pervade virtually every communicative act. Our gestures, facial expressions, postures, gaze, even physical appearance all convey messages, without anything being said. Despite their critical role in social life, nonverbal signals receive very limited attention as compared to the linguistic counterparts, and existing solutions typically examine nonverbal cues in isolation. Our study marks the first systematic effort to enhance the interpretation of multifaceted nonverbal signals. First, we contribute a novel large-scale dataset, called NVI, which is meticulously annotated to include bounding boxes for humans and corresponding social groups, along with 22 atomic-level nonverbal behaviors under five broad interaction types. Second, we establish a new task NVI-DET for nonverbal interaction detection, which is formalized as identifying triplets in the form from images. Third, we propose a nonverbal interaction detection hypergraph (NVI-DEHR), a new approach that explicitly models high-order nonverbal interactions using hypergraphs. Central to the model is a dual multi-scale hypergraph that adeptly addresses individual-to-individual and group-to-group correlations across varying scales, facilitating interactional feature learning and eventually improving interaction prediction. Extensive experiments on NVI show that NVI-DEHR improves various baselines significantly in NVI-DET. It also exhibits leading performance on HOI-DET, confirming its versatility in supporting related tasks and strong generalization ability. We hope that our study will offer the community new avenues to explore nonverbal signals in more depth.

cs.CV↗
arXiv2024

Shape2Scene: 3D Scene Representation Learning Through Pre-training on Shape Data

Current 3D self-supervised learning methods of 3D scenes face a data desert issue, resulting from the time-consuming and expensive collecting process of 3D scene data. Conversely, 3D shape datasets are easier to collect. Despite this, existing pre-training strategies on shape data offer limited potential for 3D scene understanding due to significant disparities in point quantities. To tackle these challenges, we propose Shape2Scene (S2S), a novel method that learns representations of large-scale 3D scenes from 3D shape data. We first design multiscale and high-resolution backbones for shape and scene level 3D tasks, i.e., MH-P (point-based) and MH-V (voxel-based). MH-P/V establishes direct paths to highresolution features that capture deep semantic information across multiple scales. This pivotal nature makes them suitable for a wide range of 3D downstream tasks that tightly rely on high-resolution features. We then employ a Shape-to-Scene strategy (S2SS) to amalgamate points from various shapes, creating a random pseudo scene (comprising multiple objects) for training data, mitigating disparities between shapes and scenes. Finally, a point-point contrastive loss (PPC) is applied for the pre-training of MH-P/V. In PPC, the inherent correspondence (i.e., point pairs) is naturally obtained in S2SS. Extensive experiments have demonstrated the transferability of 3D representations learned by MH-P/V across shape-level and scene-level 3D tasks. MH-P achieves notable performance on well-known point cloud datasets (93.8% OA on ScanObjectNN and 87.6% instance mIoU on ShapeNetPart). MH-V also achieves promising performance in 3D semantic segmentation and 3D object detection.

cs.CV↗
arXiv2024

Chat-Edit-3D: Interactive 3D Scene Editing via Text Prompts

Recent work on image content manipulation based on vision-language pre-training models has been effectively extended to text-driven 3D scene editing. However, existing schemes for 3D scene editing still exhibit certain shortcomings, hindering their further interactive design. Such schemes typically adhere to fixed input patterns, limiting users' flexibility in text input. Moreover, their editing capabilities are constrained by a single or a few 2D visual models and require intricate pipeline design to integrate these models into 3D reconstruction processes. To address the aforementioned issues, we propose a dialogue-based 3D scene editing approach, termed CE3D, which is centered around a large language model that allows for arbitrary textual input from users and interprets their intentions, subsequently facilitating the autonomous invocation of the corresponding visual expert models. Furthermore, we design a scheme utilizing Hash-Atlas to represent 3D scene views, which transfers the editing of 3D scenes onto 2D atlas images. This design achieves complete decoupling between the 2D editing and 3D reconstruction processes, enabling CE3D to flexibly integrate a wide range of existing 2D or 3D visual models without necessitating intricate fusion designs. Experimental results demonstrate that CE3D effectively integrates multiple visual models to achieve diverse editing visual effects, possessing strong scene comprehension and multi-round dialog capabilities. The code is available at https://sk-fun.fun/CE3D.

cs.CV↗
arXiv2024

General and Task-Oriented Video Segmentation

We present GvSeg, a general video segmentation framework for addressing four different video segmentation tasks (i.e., instance, semantic, panoptic, and exemplar-guided) while maintaining an identical architectural design. Currently, there is a trend towards developing general video segmentation solutions that can be applied across multiple tasks. This streamlines research endeavors and simplifies deployment. However, such a highly homogenized framework in current design, where each element maintains uniformity, could overlook the inherent diversity among different tasks and lead to suboptimal performance. To tackle this, GvSeg: i) provides a holistic disentanglement and modeling for segment targets, thoroughly examining them from the perspective of appearance, position, and shape, and on this basis, ii) reformulates the query initialization, matching and sampling strategies in alignment with the task-specific requirement. These architecture-agnostic innovations empower GvSeg to effectively address each unique task by accommodating the specific properties that characterize them. Extensive experiments on seven gold-standard benchmark datasets demonstrate that GvSeg surpasses all existing specialized/general solutions by a significant margin on four different video segmentation tasks.

cs.CV↗
arXiv2024

Enhancing Hallucination Detection through Perturbation-Based Synthetic Data Generation in System Responses

Detecting hallucinations in large language model (LLM) outputs is pivotal, yet traditional fine-tuning for this classification task is impeded by the expensive and quickly outdated annotation process, especially across numerous vertical domains and in the face of rapid LLM advancements. In this study, we introduce an approach that automatically generates both faithful and hallucinated outputs by rewriting system responses. Experimental findings demonstrate that a T5-base model, fine-tuned on our generated dataset, surpasses state-of-the-art zero-shot detectors and existing synthetic generation methods in both accuracy and latency, indicating efficacy of our approach.

cs.AI↗
arXiv2024

CQIL: Inference Latency Optimization with Concurrent Computation of Quasi-Independent Layers

The fast-growing large scale language models are delivering unprecedented performance on almost all natural language processing tasks. However, the effectiveness of large language models are reliant on an exponentially increasing number of parameters. The overwhelming computation complexity incurs a high inference latency that negatively affects user experience. Existing methods to improve inference efficiency, such as tensor parallelism and quantization, target to reduce per-layer computing latency, yet overlook the cumulative latency due to the number of layers. Recent works on reducing the cumulative latency through layer removing, however, lead to significant performance drop. Motivated by the similarity of inputs among adjacent layers, we propose to identify quasi-independent layers, which can be concurrently computed to significantly decrease inference latency. We also introduce a bypassing technique to mitigate the effect of information loss. Empirical experiments of the proposed approach on the LLaMA models confirm that Concurrent Computation of Quasi-Independent Layers (CQIL) can reduce latency by up to 48.3% on LLaMA-33B, while maintaining a close level of performance.

cs.CL↗