Search

We present RodinHD, which can generate high-fidelity 3D avatars from a portrait image. Existing methods fail to capture intricate details such as hairstyles which we tackle in this paper. We first identify an overlooked problem of catastrophic forgetting that arises when fitting triplanes sequentially on many avatars, caused by the MLP decoder sharing scheme. To overcome this issue, we raise a novel data scheduling strategy and a weight consolidation regularization term, which improves the decoder's capability of rendering sharper details. Additionally, we optimize the guiding effect of the portrait image by computing a finer-grained hierarchical representation that captures rich 2D texture cues, and injecting them to the 3D diffusion model at multiple layers via cross-attention. When trained on 46K avatars with a noise schedule optimized for triplanes, the resulting model can generate 3D avatars with notably better details than previous methods and can generalize to in-the-wild portrait input., Comment: ECCV 2024; project page: https://rodinhd.github.io/

Transformers have found broad applications for their great ability to capture long-range dependency among the inputs using attention mechanisms. The recent success of transformers increases the need for mathematical interpretation of their underlying working mechanisms, leading to the development of a family of white-box transformer-like deep network architectures. However, designing white-box transformers with efficient three-dimensional (3D) attention is still an open challenge. In this work, we revisit the 3D-orthogonal matching pursuit (OMP) algorithm and demonstrate that the operation of 3D-OMP is analogous to a specific kind of transformer with 3D attention. Therefore, we build a white-box 3D-OMP-transformer by introducing additional learnable parameters to 3D-OMP. As a transformer, its 3D-attention can be mathematically interpreted from 3D-OMP; while as a variant of OMP, it can learn to improve the matching pursuit process from data. Besides, a transformer's performance can be improved by stacking more transformer blocks. To simulate this process, we design a cascaded 3D-OMP-Transformer with dynamic small-scale dictionaries, which can improve the performance of the 3D-OMP-Transformer with low costs. We evaluate the designed 3D-OMP-transformer in the multi-target detection task of integrated sensing and communications (ISAC). Experimental results show that the designed 3D-OMP-Transformer can outperform current baselines.

Graph classification is a pivotal challenge in machine learning, especially within the realm of graph-based data, given its importance in numerous real-world applications such as social network analysis, recommendation systems, and bioinformatics. Despite its significance, graph classification faces several hurdles, including adapting to diverse prediction tasks, training across multiple target domains, and handling small-sample prediction scenarios. Current methods often tackle these challenges individually, leading to fragmented solutions that lack a holistic approach to the overarching problem. In this paper, we propose an algorithm aimed at addressing the aforementioned challenges. By incorporating insights from various types of tasks, our method aims to enhance adaptability, scalability, and generalizability in graph classification. Motivated by the recognition that the underlying subgraph plays a crucial role in GNN prediction, while the remainder is task-irrelevant, we introduce the Core Knowledge Learning (\method{}) framework for graph adaptation and scalability learning. \method{} comprises several key modules, including the core subgraph knowledge submodule, graph domain adaptation module, and few-shot learning module for downstream tasks. Each module is tailored to tackle specific challenges in graph classification, such as domain shift, label inconsistencies, and data scarcity. By learning the core subgraph of the entire graph, we focus on the most pertinent features for task relevance. Consequently, our method offers benefits such as improved model performance, increased domain adaptability, and enhanced robustness to domain variations. Experimental results demonstrate significant performance enhancements achieved by our method compared to state-of-the-art approaches.

Unpaired medical image synthesis aims to provide complementary information for an accurate clinical diagnostics, and address challenges in obtaining aligned multi-modal medical scans. Transformer-based models excel in imaging translation tasks thanks to their ability to capture long-range dependencies. Although effective in supervised training settings, their performance falters in unpaired image synthesis, particularly in synthesizing structural details. This paper empirically demonstrates that, lacking strong inductive biases, Transformer can converge to non-optimal solutions in the absence of paired data. To address this, we introduce UNet Structured Transformer (UNest), a novel architecture incorporating structural inductive biases for unpaired medical image synthesis. We leverage the foundational Segment-Anything Model to precisely extract the foreground structure and perform structural attention within the main anatomy. This guides the model to learn key anatomical regions, thus improving structural synthesis under the lack of supervision in unpaired training. Evaluated on two public datasets, spanning three modalities, i.e., MR, CT, and PET, UNest improves recent methods by up to 19.30% across six medical image synthesis tasks. Our code is released at https://github.com/HieuPhan33/MICCAI2024-UNest., Comment: MICCAI2024 - Early Accept Top 11%

Foundation models have become prominent in computer vision, achieving notable success in various tasks. However, their effectiveness largely depends on pre-training with extensive datasets. Applying foundation models directly to small datasets of capsule endoscopy images from scratch is challenging. Pre-training on broad, general vision datasets is crucial for successfully fine-tuning our model for specific tasks. In this work, we introduce a simplified approach called Adapt foundation models with a low-rank adaptation (LoRA) technique for easier customization. Our method, inspired by the DINOv2 foundation model, applies low-rank adaptation learning to tailor foundation models for capsule endoscopy diagnosis effectively. Unlike traditional fine-tuning methods, our strategy includes LoRA layers designed to absorb specific surgical domain knowledge. During the training process, we keep the main model (the backbone encoder) fixed and focus on optimizing the LoRA layers and the disease classification component. We tested our method on two publicly available datasets for capsule endoscopy disease classification. The results were impressive, with our model achieving 97.75% accuracy on the Kvasir-Capsule dataset and 98.81% on the Kvasirv2 dataset. Our solution demonstrates that foundation models can be adeptly adapted for capsule endoscopy diagnosis, highlighting that mere reliance on straightforward fine-tuning or pre-trained models from general computer vision tasks is inadequate for such specific applications., Comment: To appear in ICBIR 2024

Semantic Textual Similarity (STS) constitutes a critical research direction in computational linguistics and serves as a key indicator of the encoding capabilities of embedding models. Driven by advances in pre-trained language models and contrastive learning techniques, leading sentence representation methods can already achieved average Spearman's correlation scores of approximately 86 across seven STS benchmarks in SentEval. However, further improvements have become increasingly marginal, with no existing method attaining an average score higher than 87 on these tasks. This paper conducts an in-depth analysis of this phenomenon and concludes that the upper limit for Spearman's correlation scores using contrastive learning is 87.5. To transcend this ceiling, we propose an innovative approach termed Pcc-tuning, which employs Pearson's correlation coefficient as a loss function to refine model performance beyond contrastive learning. Experimental results demonstrate that Pcc-tuning markedly surpasses previous state-of-the-art strategies, raising the Spearman's correlation score to above 90., Comment: Work in Progress

Fabrication imperfections must be considered during configuration to ensure that the setup is suitable for the actual fabricated programmable photonic integrated circuits (PPICs). Therefore, characterization of imperfections is crucial but difficult, especially for PPICs made from recirculating waveguide meshes. The flexibility required by these meshes demands a more complex topology and compact TBU structure, complicating the characterization. In this paper, we propose a characterization method applicable to recirculating waveguide meshes based on an optimization approach, along with a step-by-step procedure to reduce the parameter space of optimization, allowing for characterizing imperfect parameters of each individual component within the waveguide mesh. To the best of our knowledge, this method can greatly broaden the range of characterized parameters compared to currently reported methods. In order to verify the effectiveness of our method, we used the characterized parameters to build a multi-frequency model of a mesh with fabrication errors and successfully demonstrated accurate prediction of its behavior. Furthermore, we applied our method on implementations of 6 different kind of FIR/IRR filters, to further prove the effectiveness of our method in configuring applications on meshes with fabrication errors. At last, our method was carried out under various scenarios considering beam splitter splitting ratio variance, inaccurate measurements of mesh and imprecise TBU insertion loss characterization, to demonstrate its strong robustness under various practical scenarios.

Since the introduction of BERT and RoBERTa, research on Semantic Textual Similarity (STS) has made groundbreaking progress. Particularly, the adoption of contrastive learning has substantially elevated state-of-the-art performance across various STS benchmarks. However, contrastive learning categorizes text pairs as either semantically similar or dissimilar, failing to leverage fine-grained annotated information and necessitating large batch sizes to prevent model collapse. These constraints pose challenges for researchers engaged in STS tasks that require nuanced similarity levels or those with limited computational resources, compelling them to explore alternatives like Sentence-BERT. Nonetheless, Sentence-BERT tackles STS tasks from a classification perspective, overlooking the progressive nature of semantic relationships, which results in suboptimal performance. To bridge this gap, this paper presents an innovative regression framework and proposes two simple yet effective loss functions: Translated ReLU and Smooth K2 Loss. Experimental analyses demonstrate that our method achieves convincing performance across seven established STS benchmarks, especially when supplemented with task-specific training data., Comment: Work in Progress

We search for an optimal filter design for the estimation of stellar metallicity, based on synthetic photometry from Gaia XP spectra convolved with a series of filter-transmission curves defined by different central wavelengths and bandwidths. Unlike previous designs based solely on maximizing metallicity sensitivity, we find that the optimal solution provides a balance between the sensitivity and uncertainty of the spectra. With this optimal filter design, the best precision of metallicity estimates for relatively bright ($G \sim 11.5$) stars is excellent, $\sigma_{\rm [Fe/H]} = 0.034$\,dex for FGK dwarf stars, superior to that obtained utilizing custom sensitivity-optimized filters (e.g., SkyMapper\,$v$). By selecting hundreds of high-probabability member stars of the open cluster M67, our analysis reveals that the intrinsic photometric-metallicity scatter of these cluster members is only 0.036\,dex, consistent with this level of precision. Our results clearly demonstrate that the internal precision of photometric-metallicity estimates can be extremely high, even providing the opportunity to perform chemical tagging for very large numbers of field stars in the Milky Way. This experiment shows that it is crucial to take into account uncertainty alongside the sensitivity when designing filters for measuring the stellar metallicity and other parameters., Comment: 9 pages, 5 figures; ApJL accepted, see main result in Figures 5

Diffusion-based video generation has achieved significant progress, yet generating multiple actions that occur sequentially remains a formidable task. Directly generating a video with sequential actions can be extremely challenging due to the scarcity of fine-grained action annotations and the difficulty in establishing temporal semantic correspondences and maintaining long-term consistency. To tackle this, we propose an intuitive and straightforward solution: splicing multiple single-action video segments sequentially. The core challenge lies in generating smooth and natural transitions between these segments given the inherent complexity and variability of action transitions. We introduce MAVIN (Multi-Action Video INfilling model), designed to generate transition videos that seamlessly connect two given videos, forming a cohesive integrated sequence. MAVIN incorporates several innovative techniques to address challenges in the transition video infilling task. Firstly, a consecutive noising strategy coupled with variable-length sampling is employed to handle large infilling gaps and varied generation lengths. Secondly, boundary frame guidance (BFG) is proposed to address the lack of semantic guidance during transition generation. Lastly, a Gaussian filter mixer (GFM) dynamically manages noise initialization during inference, mitigating train-test discrepancy while preserving generation flexibility. Additionally, we introduce a new metric, CLIP-RS (CLIP Relative Smoothness), to evaluate temporal coherence and smoothness, complementing traditional quality-based metrics. Experimental results on horse and tiger scenarios demonstrate MAVIN's superior performance in generating smooth and coherent video transitions compared to existing methods.

The proposed research aims to develop an innovative semantic query processing system that enables users to obtain comprehensive information about research works produced by Computer Science (CS) researchers at the Australian National University (ANU). The system integrates Large Language Models (LLMs) with the ANU Scholarly Knowledge Graph (ASKG), a structured repository of all research-related artifacts produced at ANU in the CS field. Each artifact and its parts are represented as textual nodes stored in a Knowledge Graph (KG). To address the limitations of traditional scholarly KG construction and utilization methods, which often fail to capture fine-grained details, we propose a novel framework that integrates the Deep Document Model (DDM) for comprehensive document representation and the KG-enhanced Query Processing (KGQP) for optimized complex query handling. DDM enables a fine-grained representation of the hierarchical structure and semantic relationships within academic papers, while KGQP leverages the KG structure to improve query accuracy and efficiency with LLMs. By combining the ASKG with LLMs, our approach enhances knowledge utilization and natural language understanding capabilities. The proposed system employs an automatic LLM-SPARQL fusion to retrieve relevant facts and textual nodes from the ASKG. Initial experiments demonstrate that our framework is superior to baseline methods in terms of accuracy retrieval and query efficiency. We showcase the practical application of our framework in academic research scenarios, highlighting its potential to revolutionize scholarly knowledge management and discovery. This work empowers researchers to acquire and utilize knowledge from documents more effectively and provides a foundation for developing precise and reliable interactions with LLMs., Comment: for the associated repository, see http://w3id.org/kgcp/KGQP

Static analysis techniques enhance the security, performance, and reliability of programs by analyzing and portraiting program behaviors without the need for actual execution. In essence, static analysis takes the Intermediate Representation (IR) of a target program as input to retrieve essential program information and understand the program. However, there is a lack of systematic analysis on the benefit of IR for static analysis, besides serving as an information provider. In general, a modern static analysis framework should possess the ability to conduct diverse analyses on different languages, producing reliable results with minimal time consumption, and offering extensive customization options. In this survey, we systematically characterize these goals and review the potential solutions from the perspective of IR. It can serve as a manual for learners and practitioners in the static analysis field to better understand IR design. Meanwhile, numerous research opportunities are revealed for researchers.

This paper introduces ReeSPOT, a novel Reeb graph-based method to model patterns of life in human trajectories (akin to a fingerprint). Human behavior typically follows a pattern of normalcy in day-to-day activities. This is marked by recurring activities within specific time periods. In this paper, we model this behavior using Reeb graphs where any deviation from usual day-to-day activities is encoded as nodes in the Reeb graph. The complexity of the proposed algorithm is linear with respect to the number of time points in a given trajectory. We demonstrate the usage of ReeSPOT and how it captures the critically significant spatial and temporal deviations using the nodes of the Reeb graph. Our case study presented in this paper includes realistic human movement scenarios: visiting uncommon locations, taking odd routes at infrequent times, uncommon time visits, and uncommon stay durations. We analyze the Reeb graph to interpret the topological structure of the GPS trajectories. Potential applications of ReeSPOT include urban planning, security surveillance, and behavioral research.

Conductive metal-organic frameworks (MOFs) are emerging electroactive materials for (opto-)electronics. However, it remains a great challenge to achieve reliable MOF-based devices via the existing synthesis methods that are compatible with the complementary metal-oxide-semiconductor technology, as the surface roughness of thus-far synthetic MOF films or pellets is rather high for efficient electrode contact. Here, we develop an on-liquid-gallium surface synthesis (OLGSS) strategy under chemical vapor deposition (CVD) conditions for the controlled growth of two-dimensional conjugated MOF (2D c-MOF) thin films with ten-fold improvement of surface flatness (surface roughness can reach as low as ~2 {\AA}) compared with MOF films grown by the traditional methods. Supported by theoretical modeling, we unveil a layer-by-layer CVD growth mode for constructing flattening surfaces, that is triggered by the high adhesion energy between gallium (Ga) and planar aromatic ligands. We further demonstrate the generality of the as-proposed OLGSS strategy by reproducing such a flat surface over nine different 2D c-MOF films with variable thicknesses (~2 to 208 nm) and large lateral sizes (over 1 cm2). The resultant ultra-smooth 2D c-MOF films enable the formation of high-quality electrical contacts with gold (Au) electrodes, leading to a reduction of contact resistance by over ten orders of magnitude compared to the traditional uneven MOF films. Furthermore, due to the efficient interfacial interaction benifited from the high-quality contacts, the prepared van der Waals heterostructure (vdWH) of OLGSS c-MOF and MoS2 exhibits intriguing photoluminescence (PL) enhancement, PL peak shift and large work function modulation. The establishment of the reliable OLGSS method provides the chances to push the development of MOF electronics and the construction of multicomponent MOF-based heterostructure materials.

In the landscape of spatio-temporal data analytics, effective trajectory representation learning is paramount. To bridge the gap of learning accurate representations with efficient and flexible mechanisms, we introduce Efflex, a comprehensive pipeline for transformative graph modeling and representation learning of the large-volume spatio-temporal trajectories. Efflex pioneers the incorporation of a multi-scale k-nearest neighbors (KNN) algorithm with feature fusion for graph construction, marking a leap in dimensionality reduction techniques by preserving essential data features. Moreover, the groundbreaking graph construction mechanism and the high-performance lightweight GCN increase embedding extraction speed by up to 36 times faster. We further offer Efflex in two versions, Efflex-L for scenarios demanding high accuracy, and Efflex-B for environments requiring swift data processing. Comprehensive experimentation with the Porto and Geolife datasets validates our approach, positioning Efflex as the state-of-the-art in the domain. Such enhancements in speed and accuracy highlight the versatility of Efflex, underscoring its wide-ranging potential for deployment in time-sensitive and computationally constrained applications.

IW And-type dwarf novae are anomalous Z Cam stars featured with outbursts happening during standstill states, which are not expected in the standard disk instability model. The physical mechanisms for these variations remain unclear. In this study, we report the discovery of a new candidate IW And-type dwarf nova J0652+2436, identified with its frequent outbursts from the slowly rising standstill states. Luckily, the TESS observations during a long standstill state and the earlier K2 observations give a chance to find the orbital and negative superhump period in the light curve of J0652+2436, allowing the measurement of its mass ratio of 0.366. This mass ratio is marginally possible for the tidal instability to set in according to previous SPH simulations. Thus, we propose that the outbursts in J0652+2436 are likely to be caused by the growing accretion disk during standstills, in favor of the previous hypothesis of the mechanisms lying in all IW And stars. We conclude that J0652+2436 might be the first IW And star with both a precessing tilted disk and tidal instability, which will be an important laboratory for studying the accretion disk dynamics and help understand IW And phenomenon., Comment: 12 pages, 12 figures, accepted to MNRAS

Trajectory similarity search plays an essential role in autonomous driving, as it enables vehicles to analyze the information and characteristics of different trajectories to make informed decisions and navigate safely in dynamic environments. Existing work on the trajectory similarity search task primarily utilizes sequence-processing algorithms or Recurrent Neural Networks (RNNs), which suffer from the inevitable issues of complicated architecture and heavy training costs. Considering the intricate connections between trajectories, using Graph Neural Networks (GNNs) for data modeling is feasible. However, most methods directly use existing mathematical graph structures as the input instead of constructing specific graphs from certain vehicle trajectory data. This ignores such data's unique and dynamic characteristics. To bridge such a research gap, we propose VeTraSS -- an end-to-end pipeline for Vehicle Trajectory Similarity Search. Specifically, VeTraSS models the original trajectory data into multi-scale graphs, and generates comprehensive embeddings through a novel multi-layer attention-based GNN. The learned embeddings can be used for searching similar vehicle trajectories. Extensive experiments on the Porto and Geolife datasets demonstrate the effectiveness of VeTraSS, where our model outperforms existing work and reaches the state-of-the-art. This demonstrates the potential of VeTraSS for trajectory analysis and safe navigation in self-driving vehicles in the real world.

While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing., Comment: Preprint. 14 pages, 4 figures

With the rapid development of Kesterite Cu2ZnSn(S, Se)4 solar cells in the past few years, how to achieve higher cost-performance ratio has become an important topic in the future development and industrialization of this technology. Herein, we demonstrate an all-solution route for the cell fabrication, in particular targeting at the solution processed window layer comprised of ZnO nanoparticles/Ag nanowires. A multi-interface engineering strategy assisted by organic polymers and molecules is explored to synergistically improve the film deposition, passivate the surface defects and facilitate the charge transfer. These efforts help us achieve high-performance and robust Kesterite solar cells at extremely low time and energy costs, with efficiency records of 14.37% and 13.12% being realized in rigid and flexible Kesterite solar cells, respectively. Our strategy here is also promising to be transplanted into other solar cells with similar geometric and energy band structures, helping reduce production costs and shorten the production cycle (i.e. increasing production capacity) of these photovoltaic industries.

Atomic disorder, a widespread problem in compound crystalline materials, is a imperative affecting the performance of multi-chalcogenide Cu2ZnSn(S, Se)4 (CZTSSe) photovoltaic device known for its low cost and environmental friendliness. Cu-Zn disorder is particularly abundantly present in CZTSSe due to its extraordinarily low formation energy, having induced high-concentration deep defects and severe charge loss, while its regulation remains challenging due to the contradiction between disorder-order phase transition thermodynamics and atom-interchange kinetics. Herein, through introducing more vacancies in the CZTSSe surface, we explored a vacancy-assisted strategy to reduce the atom-interchange barrier limit to facilitate the Cu-Zn ordering kinetic process. The improvement in the Cu-Zn order degree has significantly reduced the charge loss in the device and helped us realize 15.4% (certified at 14.9%) and 13.5% efficiency (certified at 13.3%) in 0.27 cm2 and 1.1 cm2-area CZTSSe solar cells, respectively, thus bringing substantial advancement for emerging inorganic thin-film photovoltaics.

Sentence Embedding stands as a fundamental task within the realm of Natural Language Processing, finding extensive application in search engines, expert systems, and question-and-answer platforms. With the continuous evolution of large language models such as LLaMA and Mistral, research on sentence embedding has recently achieved notable breakthroughs. However, these advancements mainly pertain to fine-tuning scenarios, leaving explorations into computationally efficient direct inference methods for sentence representation in a nascent stage. This paper endeavors to bridge this research gap. Through comprehensive experimentation, we challenge the widely held belief in the necessity of an Explicit One-word Limitation for deriving sentence embeddings from Pre-trained Language Models (PLMs). We demonstrate that this approach, while beneficial for generative models under direct inference scenario, is not imperative for discriminative models or the fine-tuning of generative PLMs. This discovery sheds new light on the design of manual templates in future studies. Building upon this insight, we propose two innovative prompt engineering techniques capable of further enhancing the expressive power of PLMs' raw embeddings: Pretended Chain of Thought and Knowledge Enhancement. We confirm their effectiveness across various PLM types and provide a detailed exploration of the underlying factors contributing to their success., Comment: Accepted by ICIC 2024 (Oral)

In this work, we are interested in automated methods for knowledge graph creation (KGC) from input text. Progress on large language models (LLMs) has prompted a series of recent works applying them to KGC, e.g., via zero/few-shot prompting. Despite successes on small domain-specific datasets, these models face difficulties scaling up to text common in many real-world applications. A principal issue is that in prior methods, the KG schema has to be included in the LLM prompt to generate valid triplets; larger and more complex schema easily exceed the LLMs' context window length. To address this problem, we propose a three-phase framework named Extract-Define-Canonicalize (EDC): open information extraction followed by schema definition and post-hoc canonicalization. EDC is flexible in that it can be applied to settings where a pre-defined target schema is available and when it is not; in the latter case, it constructs a schema automatically and applies self-canonicalization. To further improve performance, we introduce a trained component that retrieves schema elements relevant to the input text; this improves the LLMs' extraction performance in a retrieval-augmented generation-like manner. We demonstrate on three KGC benchmarks that EDC is able to extract high-quality triplets without any parameter tuning and with significantly larger schemas compared to prior works., Comment: 15 pages, 2 figures

This paper presents our winning submission to Subtask 2 of SemEval 2024 Task 3 on multimodal emotion cause analysis in conversations. We propose a novel Multimodal Emotion Recognition and Multimodal Emotion Cause Extraction (MER-MCE) framework that integrates text, audio, and visual modalities using specialized emotion encoders. Our approach sets itself apart from top-performing teams by leveraging modality-specific features for enhanced emotion understanding and causality inference. Experimental evaluation demonstrates the advantages of our multimodal approach, with our submission achieving a competitive weighted F1 score of 0.3435, ranking third with a margin of only 0.0339 behind the 1st team and 0.0025 behind the 2nd team. Project: https://github.com/MIPS-COLT/MER-MCE.git, Comment: Ranked 3rd in SemEval '24 Task 3 with F1 of 0.3435, close to 1st & 2nd by 0.0339 & 0.0025

Sulfide Kesterite Cu2ZnSnS4 (CZTS), a nontoxic and low-cost photovoltaic material, has always being facing severe charge recombination and poor carrier transport, resulting in the cell efficiency record stagnating around 11% for years. Gradient bandgap is a promising approach to relieve these issues, however, has not been effectively realized in Kesterite solar cells due to the challenges in controlling the gradient distribution of alloying elements at high temperatures. Herein, targeting at the Cd alloyed CZTS, we propose a pre-crystallization strategy to reduce the intense vertical mass transport and Cd rapid diffusion in the film growth process, thereby realizing front Cd-gradient CZTS absorber. The Cd-gradient CZTS absorber, exhibiting downward bending conduction band structure, has significantly enhanced the minority carrier transport and additionally improved band alignment and interface property of CZTS/CdS heterojunction. Ultimately, we have achieved a champion total-area efficiency of 13.5% (active-area efficiency: 14.1%) in the cell and in particular a high open-circuit voltage of >800 mV. We have also achieved a certified total-area cell efficiency of 13.16%, realizing a substantial step forward for the pure sulfide Kesterite solar cell.

We introduce a radiance representation that is both structured and fully explicit and thus greatly facilitates 3D generative modeling. Existing radiance representations either require an implicit feature decoder, which significantly degrades the modeling power of the representation, or are spatially unstructured, making them difficult to integrate with mainstream 3D diffusion methods. We derive GaussianCube by first using a novel densification-constrained Gaussian fitting algorithm, which yields high-accuracy fitting using a fixed number of free Gaussians, and then rearranging these Gaussians into a predefined voxel grid via Optimal Transport. Since GaussianCube is a structured grid representation, it allows us to use standard 3D U-Net as our backbone in diffusion modeling without elaborate designs. More importantly, the high-accuracy fitting of the Gaussians allows us to achieve a high-quality representation with orders of magnitude fewer parameters than previous structured representations for comparable quality, ranging from one to two orders of magnitude. The compactness of GaussianCube greatly eases the difficulty of 3D generative modeling. Extensive experiments conducted on unconditional and class-conditioned object generation, digital avatar creation, and text-to-3D synthesis all show that our model achieves state-of-the-art generation results both qualitatively and quantitatively, underscoring the potential of GaussianCube as a highly accurate and versatile radiance representation for 3D generative modeling. Project page: https://gaussiancube.github.io/., Comment: Update for digital avatar creation and text-to-3D synthesis; Project Page: https://gaussiancube.github.io/

Stance detection aims to determine the attitude expressed in text towards a given target. Zero-shot stance detection (ZSSD) has emerged to classify stances towards unseen targets during inference. Recent data augmentation techniques for ZSSD increase transferable knowledge between targets through text or target augmentation. However, these methods exhibit limitations. Target augmentation lacks logical connections between generated targets and source text, while text augmentation relies solely on training data, resulting in insufficient generalization. To address these issues, we propose an encoder-decoder data augmentation (EDDA) framework. The encoder leverages large language models and chain-of-thought prompting to summarize texts into target-specific if-then rationales, establishing logical relationships. The decoder generates new samples based on these expressions using a semantic correlation word replacement strategy to increase syntactic diversity. We also analyze the generated expressions to develop a rationale-enhanced network that fully utilizes the augmented data. Experiments on benchmark datasets demonstrate our approach substantially improves over state-of-the-art ZSSD techniques. The proposed EDDA framework increases semantic relevance and syntactic variety in augmented texts while enabling interpretable rationale-based learning.

3D generation has witnessed significant advancements, yet efficiently producing high-quality 3D assets from a single image remains challenging. In this paper, we present a triplane autoencoder, which encodes 3D models into a compact triplane latent space to effectively compress both the 3D geometry and texture information. Within the autoencoder framework, we introduce a 3D-aware cross-attention mechanism, which utilizes low-resolution latent representations to query features from a high-resolution 3D feature volume, thereby enhancing the representation capacity of the latent space. Subsequently, we train a diffusion model on this refined latent space. In contrast to solely relying on image embedding for 3D generation, our proposed method advocates for the simultaneous utilization of both image embedding and shape embedding as conditions. Specifically, the shape embedding is estimated via a diffusion prior model conditioned on the image embedding. Through comprehensive experiments, we demonstrate that our method outperforms state-of-the-art algorithms, achieving superior performance while requiring less training data and time. Our approach enables the generation of high-quality 3D assets in merely 7 seconds on a single A100 GPU.

Previous stance detection studies typically concentrate on evaluating stances within individual instances, thereby exhibiting limitations in effectively modeling multi-party discussions concerning the same specific topic, as naturally transpire in authentic social media interactions. This constraint arises primarily due to the scarcity of datasets that authentically replicate real social media contexts, hindering the research progress of conversational stance detection. In this paper, we introduce a new multi-turn conversation stance detection dataset (called \textbf{MT-CSD}), which encompasses multiple targets for conversational stance detection. To derive stances from this challenging dataset, we propose a global-local attention network (\textbf{GLAN}) to address both long and short-range dependencies inherent in conversational data. Notably, even state-of-the-art stance detection methods, exemplified by GLAN, exhibit an accuracy of only 50.47\%, highlighting the persistent challenges in conversational stance detection. Furthermore, our MT-CSD dataset serves as a valuable resource to catalyze advancements in cross-domain stance detection, where a classifier is adapted from a different yet related target. We believe that MT-CSD will contribute to advancing real-world applications of stance detection research. Our source code, data, and models are available at \url{https://github.com/nfq729/MT-CSD}.

In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting.

Medical vision language pre-training (VLP) has emerged as a frontier of research, enabling zero-shot pathological recognition by comparing the query image with the textual descriptions for each disease. Due to the complex semantics of biomedical texts, current methods struggle to align medical images with key pathological findings in unstructured reports. This leads to the misalignment with the target disease's textual representation. In this paper, we introduce a novel VLP framework designed to dissect disease descriptions into their fundamental aspects, leveraging prior knowledge about the visual manifestations of pathologies. This is achieved by consulting a large language model and medical experts. Integrating a Transformer module, our approach aligns an input image with the diverse elements of a disease, generating aspect-centric image representations. By consolidating the matches from each aspect, we improve the compatibility between an image and its associated disease. Additionally, capitalizing on the aspect-oriented representations, we present a dual-head Transformer tailored to process known and unknown diseases, optimizing the comprehensive detection efficacy. Conducting experiments on seven downstream datasets, ours improves the accuracy of recent methods by up to 8.56% and 17.26% for seen and unseen categories, respectively. Our code is released at https://github.com/HieuPhan33/MAVL., Comment: Accepted at CVPR2024. Pre-print before final camera-ready version

Token-based text-to-speech (TTS) models have emerged as a promising avenue for generating natural and realistic speech, yet they grapple with low pronunciation accuracy, speaking style and timbre inconsistency, and a substantial need for diverse training data. In response, we introduce a novel hierarchical acoustic modeling approach complemented by a tailored data augmentation strategy and train it on the combination of real and synthetic data, scaling the data size up to 650k hours, leading to the zero-shot TTS model with 0.8B parameters. Specifically, our method incorporates a latent variable sequence containing supplementary acoustic information based on refined self-supervised learning (SSL) discrete units into the TTS model by a predictor. This significantly mitigates pronunciation errors and style mutations in synthesized speech. During training, we strategically replace and duplicate segments of the data to enhance timbre uniformity. Moreover, a pretrained few-shot voice conversion model is utilized to generate a plethora of voices with identical content yet varied timbres. This facilitates the explicit learning of utterance-level one-to-many mappings, enriching speech diversity and also ensuring consistency in timbre. Comparative experiments (Demo page: https://anonymous.4open.science/w/ham-tts/)demonstrate our model's superiority over VALL-E in pronunciation precision and maintaining speaking style, as well as timbre continuity.

Large language models (LLMs) have demonstrated impressive performance in various natural language processing (NLP) tasks. However, there is limited understanding of how well LLMs perform in specific domains (e.g, the intellectual property (IP) domain). In this paper, we contribute a new benchmark, the first Multilingual-oriented quiZ on Intellectual Property (MoZIP), for the evaluation of LLMs in the IP domain. The MoZIP benchmark includes three challenging tasks: IP multiple-choice quiz (IPQuiz), IP question answering (IPQA), and patent matching (PatentMatch). In addition, we also develop a new IP-oriented multilingual large language model (called MoZi), which is a BLOOMZ-based model that has been supervised fine-tuned with multilingual IP-related text data. We evaluate our proposed MoZi model and four well-known LLMs (i.e., BLOOMZ, BELLE, ChatGLM and ChatGPT) on the MoZIP benchmark. Experimental results demonstrate that MoZi outperforms BLOOMZ, BELLE and ChatGLM by a noticeable margin, while it had lower scores compared with ChatGPT. Notably, the performance of current LLMs on the MoZIP benchmark has much room for improvement, and even the most powerful ChatGPT does not reach the passing level. Our source code, data, and models are available at \url{https://github.com/AI-for-Science/MoZi}.

Source-Free Unsupervised Domain Adaptation (SFUDA) is a challenging task where a model needs to be adapted to a new domain without access to target domain labels or source domain data. The primary difficulty in this task is that the model's predictions may be inaccurate, and using these inaccurate predictions for model adaptation can lead to misleading results. To address this issue, this paper proposes a novel approach that considers multiple prediction hypotheses for each sample and investigates the rationale behind each hypothesis. By consolidating these hypothesis rationales, we identify the most likely correct hypotheses, which we then use as a pseudo-labeled set to support a semi-supervised learning procedure for model adaptation. To achieve the optimal performance, we propose a three-step adaptation process: model pre-adaptation, hypothesis consolidation, and semi-supervised learning. Extensive experimental results demonstrate that our approach achieves state-of-the-art performance in the SFUDA task and can be easily integrated into existing approaches to improve their performance. The codes are available at \url{https://github.com/GANPerf/HCPR}.

Indoor imaging is a critical task for robotics and internet-of-things. WiFi as an omnipresent signal is a promising candidate for carrying out passive imaging and synchronizing the up-to-date information to all connected devices. This is the first research work to consider WiFi indoor imaging as a multi-modal image generation task that converts the measured WiFi power into a high-resolution indoor image. Our proposed WiFi-GEN network achieves a shape reconstruction accuracy that is 275% of that achieved by physical model-based inversion methods. Additionally, the Frechet Inception Distance score has been significantly reduced by 82%. To examine the effectiveness of models for this task, the first large-scale dataset is released containing 80,000 pairs of WiFi signal and imaging target. Our model absorbs challenges for the model-based methods including the non-linearity, ill-posedness and non-certainty into massive parameters of our generative AI network. The network is also designed to best fit measured WiFi signals and the desired imaging output. For reproducibility, we will release the data and code upon acceptance.

Cross-target stance detection (CTSD) is an important task, which infers the attitude of the destination target by utilizing annotated data derived from the source target. One important approach in CTSD is to extract domain-invariant features to bridge the knowledge gap between multiple targets. However, the analysis of informal and short text structure, and implicit expressions, complicate the extraction of domain-invariant knowledge. In this paper, we propose a Multi-Perspective Prompt-Tuning (MPPT) model for CTSD that uses the analysis perspective as a bridge to transfer knowledge. First, we develop a two-stage instruct-based chain-of-thought method (TsCoT) to elicit target analysis perspectives and provide natural language explanations (NLEs) from multiple viewpoints by formulating instructions based on large language model (LLM). Second, we propose a multi-perspective prompt-tuning framework (MultiPLN) to fuse the NLEs into the stance predictor. Extensive experiments results demonstrate the superiority of MPPT against the state-of-the-art baseline methods.

Zero-shot stance detection (ZSSD) aims to detect stances toward unseen targets. Incorporating background knowledge to enhance transferability between seen and unseen targets constitutes the primary approach of ZSSD. However, these methods often struggle with a knowledge-task disconnect and lack logical consistency in their predictions. To address these issues, we introduce a novel approach named Logically Consistent Chain-of-Thought (LC-CoT) for ZSSD, which improves stance detection by ensuring relevant and logically sound knowledge extraction. LC-CoT employs a three-step process. Initially, it assesses whether supplementary external knowledge is necessary. Subsequently, it uses API calls to retrieve this knowledge, which can be processed by a separate LLM. Finally, a manual exemplar guides the LLM to infer stance categories, using an if-then logical structure to maintain relevance and logical coherence. This structured approach to eliciting background knowledge enhances the model's capability, outperforming traditional supervised methods without relying on labeled data.

Organ at risk (OAR) segmentation is a critical process in radiotherapy treatment planning such as head and neck tumors. Nevertheless, in clinical practice, radiation oncologists predominantly perform OAR segmentations manually on CT scans. This manual process is highly time-consuming and expensive, limiting the number of patients who can receive timely radiotherapy. Additionally, CT scans offer lower soft-tissue contrast compared to MRI. Despite MRI providing superior soft-tissue visualization, its time-consuming nature makes it infeasible for real-time treatment planning. To address these challenges, we propose a method called SegReg, which utilizes Elastic Symmetric Normalization for registering MRI to perform OAR segmentation. SegReg outperforms the CT-only baseline by 16.78% in mDSC and 18.77% in mIoU, showing that it effectively combines the geometric accuracy of CT with the superior soft-tissue contrast of MRI, making accurate automated OAR segmentation for clinical practice become possible. See project website https://steve-zeyu-zhang.github.io/SegReg, Comment: Accepted to ISBI 2024

Recent efforts to enable visual navigation using large language models have mainly focused on developing complex prompt systems. These systems incorporate instructions, observations, and history into massive text prompts, which are then combined with pre-trained large language models to facilitate visual navigation. In contrast, our approach aims to fine-tune large language models for visual navigation without extensive prompt engineering. Our design involves a simple text prompt, current observations, and a history collector model that gathers information from previous observations as input. For output, our design provides a probability distribution of possible actions that the agent can take during navigation. We train our model using human demonstrations and collision signals from the Habitat-Matterport 3D Dataset (HM3D). Experimental results demonstrate that our method outperforms state-of-the-art behavior cloning methods and effectively reduces collision rates.

We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret, Comment: 30 pages, 10 figures. Code/Project Website: https://github.com/apple/ml-ferret

Large-scale web-crawled datasets are fundamental for the success of pre-training vision-language models, such as CLIP. However, the inherent noise and potential irrelevance of web-crawled AltTexts pose challenges in achieving precise image-text alignment. Existing methods utilizing large language models (LLMs) for caption rewriting have shown promise on small, curated datasets like CC3M and CC12M. This study introduces a scalable pipeline for noisy caption rewriting. Unlike recent LLM rewriting techniques, we emphasize the incorporation of visual concepts into captions, termed as Visual-enriched Captions (VeCap). To ensure data diversity, we propose a novel mixed training scheme that optimizes the utilization of AltTexts alongside newly generated VeCap. We showcase the adaptation of this method for training CLIP on large-scale web-crawled datasets, termed VeCLIP. Employing this cost-effective pipeline, we effortlessly scale our dataset up to 300 million samples named VeCap dataset. Our results show significant advantages in image-text alignment and overall model performance. For example, VeCLIP achieves up to +25.2% gain in COCO and Flickr30k retrieval tasks under the 12M setting. For data efficiency, VeCLIP achieves +3% gain while only using 14% of the data employed in the vanilla CLIP and 11% in ALIGN. We also note the VeCap data is complementary with other well curated datasets good for zero-shot classification tasks. When combining VeCap and DFN, our model can achieve strong performance on both of image-text retrieval and zero-shot classification tasks, e.g. 83.1% accuracy@1 on ImageNet zero-shot for a H/14 model. We release the pre-trained models at https://github.com/apple/ml-veclip., Comment: CV/ML

Camera sensors have been widely used in intelligent robotic systems. Developing camera sensors with high sensing efficiency has always been important to reduce the power, memory, and other related resources. Inspired by recent success on programmable sensors and deep optic methods, we design a novel video compressed sensing system with spatially-variant compression ratios, which achieves higher imaging quality than the existing snapshot compressed imaging methods with the same sensing costs. In this article, we also investigate the data transmission methods for programmable sensors, where the performance of communication systems is evaluated by the reconstructed images or videos rather than the transmission of sensor data itself. Usually, different reconstruction algorithms are designed for applications in high dynamic range imaging, video compressive sensing, or motion debluring. This task-aware property inspires a semantic communication framework for programmable sensors. In this work, a policy-gradient based reinforcement learning method is introduced to achieve the explicit trade-off between the compression (or transmission) rate and the image distortion. Numerical results show the superiority of the proposed methods over existing baselines.

Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick., Comment: Accepted to ICLR 2024

We propose a new method to reconstruct the 3D human body from RGB-D images with occlusions. The foremost challenge is the incompleteness of the RGB-D data due to occlusions between the body and the environment, leading to implausible reconstructions that suffer from severe human-scene penetration. To reconstruct a semantically and physically plausible human body, we propose to reduce the solution space based on scene information and prior knowledge. Our key idea is to constrain the solution space of the human body by considering the occluded body parts and visible body parts separately: modeling all plausible poses where the occluded body parts do not penetrate the scene, and constraining the visible body parts using depth data. Specifically, the first component is realized by a neural network that estimates the candidate region named the "free zone", a region carved out of the open space within which it is safe to search for poses of the invisible body parts without concern for penetration. The second component constrains the visible body parts using the "truncated shadow volume" of the scanned body point cloud. Furthermore, we propose to use a volume matching strategy, which yields better performance than surface matching, to match the human body with the confined region. We conducted experiments on the PROX dataset, and the results demonstrate that our method produces more accurate and plausible results compared with other methods.

Unsupervised sentence representation learning aims to transform input sentences into fixed-length vectors enriched with intricate semantic information while obviating the reliance on labeled data. Recent strides within this domain have been significantly propelled by breakthroughs in contrastive learning and prompt engineering. Despite these advancements, the field has reached a plateau, leading some researchers to incorporate external components to enhance the quality of sentence embeddings. Such integration, though beneficial, complicates solutions and inflates demands for computational resources. In response to these challenges, this paper presents CoT-BERT, an innovative method that harnesses the progressive thinking of Chain-of-Thought reasoning to tap into the latent potential of pre-trained models like BERT. Additionally, we develop an advanced contrastive learning loss function and propose a novel template denoising strategy. Rigorous experimentation demonstrates that CoT-BERT surpasses a range of well-established baselines by relying exclusively on the intrinsic strengths of pre-trained models., Comment: Accepted by ICANN 2024

Sparse Mixture-of-Experts models (MoEs) have recently gained popularity due to their ability to decouple model size from inference efficiency by only activating a small subset of the model parameters for any given input token. As such, sparse MoEs have enabled unprecedented scalability, resulting in tremendous successes across domains such as natural language processing and computer vision. In this work, we instead explore the use of sparse MoEs to scale-down Vision Transformers (ViTs) to make them more attractive for resource-constrained vision applications. To this end, we propose a simplified and mobile-friendly MoE design where entire images rather than individual patches are routed to the experts. We also propose a stable MoE training procedure that uses super-class information to guide the router. We empirically show that our sparse Mobile Vision MoEs (V-MoEs) can achieve a better trade-off between performance and efficiency than the corresponding dense ViTs. For example, for the ViT-Tiny model, our Mobile V-MoE outperforms its dense counterpart by 3.39% on ImageNet-1k. For an even smaller ViT variant with only 54M FLOPs inference cost, our MoE achieves an improvement of 4.66%.

Text-to-image synthesis has made encouraging progress and attracted lots of public attention recently. However, popular evaluation metrics in this area, like the Inception Score and Fr'echet Inception Distance, incur several issues. First of all, they cannot explicitly assess the perceptual quality of generated images and poorly reflect the semantic alignment of each text-image pair. Also, they are inefficient and need to sample thousands of images to stabilise their evaluation results. In this paper, we propose to evaluate text-to-image generation performance by directly estimating the likelihood of the generated images using a pre-trained likelihood-based text-to-image generative model, i.e., a higher likelihood indicates better perceptual quality and better text-image alignment. To prevent the likelihood of being dominated by the non-crucial part of the generated image, we propose several new designs to develop a credit assignment strategy based on the semantic and perceptual significance of the image patches. In the experiments, we evaluate the proposed metric on multiple popular text-to-image generation models and datasets in accessing both the perceptual quality and the text-image alignment. Moreover, it can successfully assess the generation ability of these models with as few as a hundred samples, making it very efficient in practice.

Aggregation of multi-stage features has been revealed to play a significant role in semantic segmentation. Unlike previous methods employing point-wise summation or concatenation for feature aggregation, this study proposes the Category Feature Transformer (CFT) that explores the flow of category embedding and transformation among multi-stage features through the prevalent multi-head attention mechanism. CFT learns unified feature embeddings for individual semantic categories from high-level features during each aggregation process and dynamically broadcasts them to high-resolution features. Integrating the proposed CFT into a typical feature pyramid structure exhibits superior performance over a broad range of backbone networks. We conduct extensive experiments on popular semantic segmentation benchmarks. Specifically, the proposed CFT obtains a compelling 55.1% mIoU with greatly reduced model parameters and computations on the challenging ADE20K dataset.

Vision transformers have achieved leading performance on various visual tasks yet still suffer from high computational complexity. The situation deteriorates in dense prediction tasks like semantic segmentation, as high-resolution inputs and outputs usually imply more tokens involved in computations. Directly removing the less attentive tokens has been discussed for the image classification task but can not be extended to semantic segmentation since a dense prediction is required for every patch. To this end, this work introduces a Dynamic Token Pruning (DToP) method based on the early exit of tokens for semantic segmentation. Motivated by the coarse-to-fine segmentation process by humans, we naturally split the widely adopted auxiliary-loss-based network architecture into several stages, where each auxiliary block grades every token's difficulty level. We can finalize the prediction of easy tokens in advance without completing the entire forward pass. Moreover, we keep $k$ highest confidence tokens for each semantic category to uphold the representative context information. Thus, computational complexity will change with the difficulty of the input, akin to the way humans do segmentation. Experiments suggest that the proposed DToP architecture reduces on average $20\% - 35\%$ of computational cost for current semantic segmentation methods based on plain vision transformers without accuracy degradation.