arXiv subjects

Haoran Zhang

Publications and source records attributed to Haoran Zhang.

At least 127 records · Page 7Linked to original sources

arXiv2024

Towards Vision Mixture of Experts for Wildlife Monitoring on the Edge

The explosion of IoT sensors in industrial, consumer and remote sensing use cases has come with unprecedented demand for computing infrastructure to transmit and to analyze petabytes of data. Concurrently, the world is slowly shifting its focus towards more sustainable computing. For these reasons, there has been a recent effort to reduce the footprint of related computing infrastructure, especially by deep learning algorithms, for advanced insight generation. The `TinyML' community is actively proposing methods to save communication bandwidth and excessive cloud storage costs while reducing algorithm inference latency and promoting data privacy. Such proposed approaches should ideally process multiple types of data, including time series, audio, satellite images, and video, near the network edge as multiple data streams has been shown to improve the discriminative ability of learning algorithms, especially for generating fine grained results. Incidentally, there has been recent work on data driven conditional computation of subnetworks that has shown real progress in using a single model to share parameters among very different types of inputs such as images and text, reducing the computation requirement of multi-tower multimodal networks. Inspired by such line of work, we explore similar per patch conditional computation for the first time for mobile vision transformers (vision only case), that will eventually be used for single-tower multimodal edge models. We evaluate the model on Cornell Sap Sucker Woods 60, a fine grained bird species discrimination dataset. Our initial experiments uses $4X$ fewer parameters compared to MobileViTV2-1.0 with a $1$% accuracy drop on the iNaturalist '21 birds test data provided as part of the SSW60 dataset.

cs.CV↗

arXiv2024

TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables

Deep models have demonstrated remarkable performance in time series forecasting. However, due to the partially-observed nature of real-world applications, solely focusing on the target of interest, so-called endogenous variables, is usually insufficient to guarantee accurate forecasting. Notably, a system is often recorded into multiple variables, where the exogenous variables can provide valuable external information for endogenous variables. Thus, unlike well-established multivariate or univariate forecasting paradigms that either treat all the variables equally or ignore exogenous information, this paper focuses on a more practical setting: time series forecasting with exogenous variables. We propose a novel approach, TimeXer, to ingest external information to enhance the forecasting of endogenous variables. With deftly designed embedding layers, TimeXer empowers the canonical Transformer with the ability to reconcile endogenous and exogenous information, where patch-wise self-attention and variate-wise cross-attention are used simultaneously. Moreover, global endogenous tokens are learned to effectively bridge the causal information underlying exogenous series into endogenous temporal patches. Experimentally, TimeXer achieves consistent state-of-the-art performance on twelve real-world forecasting benchmarks and exhibits notable generality and scalability. Code is available at this repository: https://github.com/thuml/TimeXer.

cs.LG↗

arXiv2024

BendVLM: Test-Time Debiasing of Vision-Language Embeddings

Vision-language model (VLM) embeddings have been shown to encode biases present in their training data, such as societal biases that prescribe negative characteristics to members of various racial and gender identities. VLMs are being quickly adopted for a variety of tasks ranging from few-shot classification to text-guided image generation, making debiasing VLM embeddings crucial. Debiasing approaches that fine-tune the VLM often suffer from catastrophic forgetting. On the other hand, fine-tuning-free methods typically utilize a "one-size-fits-all" approach that assumes that correlation with the spurious attribute can be explained using a single linear direction across all possible inputs. In this work, we propose Bend-VLM, a nonlinear, fine-tuning-free approach for VLM embedding debiasing that tailors the debiasing operation to each unique input. This allows for a more flexible debiasing approach. Additionally, we do not require knowledge of the set of inputs a priori to inference time, making our method more appropriate for online, open-set tasks such as retrieval and text guided image generation.

cs.CV↗

arXiv2024

The Fine Line: Navigating Large Language Model Pretraining with Down-streaming Capability Analysis

Uncovering early-stage metrics that reflect final model performance is one core principle for large-scale pretraining. The existing scaling law demonstrates the power-law correlation between pretraining loss and training flops, which serves as an important indicator of the current training state for large language models. However, this principle only focuses on the model's compression properties on the training data, resulting in an inconsistency with the ability improvements on the downstream tasks. Some follow-up works attempted to extend the scaling-law to more complex metrics (such as hyperparameters), but still lacked a comprehensive analysis of the dynamic differences among various capabilities during pretraining. To address the aforementioned limitations, this paper undertakes a comprehensive comparison of model capabilities at various pretraining intermediate checkpoints. Through this analysis, we confirm that specific downstream metrics exhibit similar training dynamics across models of different sizes, up to 67 billion parameters. In addition to our core findings, we've reproduced Amber and OpenLLaMA, releasing their intermediate checkpoints. This initiative offers valuable resources to the research community and facilitates the verification and exploration of LLM pretraining by open-source researchers. Besides, we provide empirical summaries, including performance comparisons of different models and capabilities, and tuition of key metrics for different training phases. Based on these findings, we provide a more user-friendly strategy for evaluating the optimization state, offering guidance for establishing a stable pretraining process.

cs.CL↗

arXiv2024

CMMMU: A Chinese Massive Multi-discipline Multimodal Understanding Benchmark

As the capabilities of large multimodal models (LMMs) continue to advance, evaluating the performance of LMMs emerges as an increasing need. Additionally, there is an even larger gap in evaluating the advanced knowledge and reasoning abilities of LMMs in non-English contexts such as Chinese. We introduce CMMMU, a new Chinese Massive Multi-discipline Multimodal Understanding benchmark designed to evaluate LMMs on tasks demanding college-level subject knowledge and deliberate reasoning in a Chinese context. CMMMU is inspired by and strictly follows the annotation and analysis pattern of MMMU. CMMMU includes 12k manually collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering, like its companion, MMMU. These questions span 30 subjects and comprise 39 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. CMMMU focuses on complex perception and reasoning with domain-specific knowledge in the Chinese context. We evaluate 11 open-source LLMs and one proprietary GPT-4V(ision). Even GPT-4V only achieves accuracies of 42%, indicating a large space for improvement. CMMMU will boost the community to build the next-generation LMMs towards expert artificial intelligence and promote the democratization of LMMs by providing diverse language contexts.

cs.CL↗

arXiv2024

Identifying Implicit Social Biases in Vision-Language Models

Vision-language models, like CLIP (Contrastive Language Image Pretraining), are becoming increasingly popular for a wide range of multimodal retrieval tasks. However, prior work has shown that large language and deep vision models can learn historical biases contained in their training sets, leading to perpetuation of stereotypes and potential downstream harm. In this work, we conduct a systematic analysis of the social biases that are present in CLIP, with a focus on the interaction between image and text modalities. We first propose a taxonomy of social biases called So-B-IT, which contains 374 words categorized across ten types of bias. Each type can lead to societal harm if associated with a particular demographic group. Using this taxonomy, we examine images retrieved by CLIP from a facial image dataset using each word as part of a prompt. We find that CLIP frequently displays undesirable associations between harmful words and specific demographic groups, such as retrieving mostly pictures of Middle Eastern men when asked to retrieve images of a "terrorist". Finally, we conduct an analysis of the source of such biases, by showing that the same harmful stereotypes are also present in a large image-text dataset used to train CLIP models for examples of biases that we find. Our findings highlight the importance of evaluating and addressing bias in vision-language models, and suggest the need for transparency and fairness-aware curation of large pre-training datasets.

cs.CV↗

arXiv2024

Timer: Generative Pre-trained Transformers Are Large Time Series Models

Deep learning has contributed remarkably to the advancement of time series analysis. Still, deep models can encounter performance bottlenecks in real-world data-scarce scenarios, which can be concealed due to the performance saturation with small models on current benchmarks. Meanwhile, large models have demonstrated great powers in these scenarios through large-scale pre-training. Continuous progress has been achieved with the emergence of large language models, exhibiting unprecedented abilities such as few-shot generalization, scalability, and task generality, which are however absent in small deep models. To change the status quo of training scenario-specific small models from scratch, this paper aims at the early development of large time series models (LTSM). During pre-training, we curate large-scale datasets with up to 1 billion time points, unify heterogeneous time series into single-series sequence (S3) format, and develop the GPT-style architecture toward LTSMs. To meet diverse application needs, we convert forecasting, imputation, and anomaly detection of time series into a unified generative task. The outcome of this study is a Time Series Transformer (Timer), which is generative pre-trained by next token prediction and adapted to various downstream tasks with promising capabilities as an LTSM. Code and datasets are available at: https://github.com/thuml/Large-Time-Series-Model.

cs.LG↗

arXiv2024

ING-VP: MLLMs cannot Play Easy Vision-based Games Yet

As multimodal large language models (MLLMs) continue to demonstrate increasingly competitive performance across a broad spectrum of tasks, more intricate and comprehensive benchmarks have been developed to assess these cutting-edge models. These benchmarks introduce new challenges to core capabilities such as perception, reasoning, and planning. However, existing multimodal benchmarks fall short in providing a focused evaluation of multi-step planning based on spatial relationships in images. To bridge this gap, we present ING-VP, the first INteractive Game-based Vision Planning benchmark, specifically designed to evaluate the spatial imagination and multi-step reasoning abilities of MLLMs. ING-VP features 6 distinct games, encompassing 300 levels, each with 6 unique configurations. A single model engages in over 60,000 rounds of interaction. The benchmark framework allows for multiple comparison settings, including image-text vs. text-only inputs, single-step vs. multi-step reasoning, and with-history vs. without-history conditions, offering valuable insights into the model's capabilities. We evaluated numerous state-of-the-art MLLMs, with the highest-performing model, Claude-3.5 Sonnet, achieving an average accuracy of only 3.37%, far below the anticipated standard. This work aims to provide a specialized evaluation framework to drive advancements in MLLMs' capacity for complex spatial reasoning and planning. The code is publicly available at https://github.com/Thisisus7/ING-VP.git.

cs.CL↗

arXiv2024

Tokenization Is More Than Compression

Tokenization is a foundational step in natural language processing (NLP) tasks, bridging raw text and language models. Existing tokenization approaches like Byte-Pair Encoding (BPE) originate from the field of data compression, and it has been suggested that the effectiveness of BPE stems from its ability to condense text into a relatively small number of tokens. We test the hypothesis that fewer tokens lead to better downstream performance by introducing PathPiece, a new tokenizer that segments a document's text into the minimum number of tokens for a given vocabulary. Through extensive experimentation we find this hypothesis not to be the case, casting doubt on the understanding of the reasons for effective tokenization. To examine which other factors play a role, we evaluate design decisions across all three phases of tokenization: pre-tokenization, vocabulary construction, and segmentation, offering new insights into the design of effective tokenizers. Specifically, we illustrate the importance of pre-tokenization and the benefits of using BPE to initialize vocabulary construction. We train 64 language models with varying tokenization, ranging in size from 350M to 2.4B parameters, all of which are made publicly available.

cs.CL↗

arXiv2024

Polarization-encoded quantum key distribution with a room-temperature telecom single-photon emitter

Single photon sources (SPSs) are directly applicable in quantum key distribution (QKD) because they allow the implementation of the canonical BB84 protocol. To date, QKD implementations using SPS are not widespread because of the need for cryogenic operation, or frequency conversion to a wavelength efficiently transmitted over telecommunication fibers. We report an observation of polarization-encoded QKD using a room-temperature telecom SPS based on a GaN defect. A field test over 3.5 km of deployed fiber with 4.0 dB loss yielded a secure key rate of 585.9~bps. Further testing in a 32.5 km fiber spool (attenuation of 11.2 dB), which exhibited substantially lower polarization mode dispersion, yielded a secure key rate of 50.4 bps. Both results exhibited a quantum bit error rate (QBER) of approximately 5%. These results illustrate the potential of the GaN defects for supporting polarization-encoded quantum communication.

quant-ph↗

arXiv2024

Metropolitan quantum key distribution using a GaN-based room-temperature telecommunication single-photon source

Single-photon sources (SPS) hold the potential to enhance the performance of quantum key distribution (QKD). QKD systems using SPS often require cryogenic cooling, while recent QKD attempts using SPS operating at room-temperature have failed to achieve long-distance transmission due to the SPS not operating at telecommunication wavelength. In this work, we have successfully demonstrated QKD using a room-temperature SPS at telecommunication wavelength. The SPS used in this work is based on point defects hosted by gallium nitride (GaN) thin films grown on sapphire substrates. We employed a time-bin and phase encoding scheme to perform the BB84 and reference-frame-independent QKD protocols over a 33 km fiber spool, achieving a secure key rate of $7.58\times 10^{-7}$ per pulse. Moreover, we also implemented a metropolitan QKD experiment over a 30 km deployed fiber, achieving a secure key rate of $6.06\times 10^{-8}$ per pulse. These results broaden the prospects for future use of SPS in commercial QKD applications.

quant-ph↗

arXiv2024

CLSP: High-Fidelity Contrastive Language-State Pre-training for Agent State Representation

With the rapid development of artificial intelligence, multimodal learning has become an important research area. For intelligent agents, the state is a crucial modality to convey precise information alongside common modalities like images, videos, and language. This becomes especially clear with the broad adoption of reinforcement learning and multimodal large language models. Nevertheless, the representation of state modality still lags in development. To this end, we propose a High-Fidelity Contrastive Language-State Pre-training (CLSP) method, which can accurately encode state information into general representations for both reinforcement learning and multimodal large language models. Specifically, we first design a pre-training task based on the classification to train an encoder with coarse-grained information. Next, we construct data pairs of states and language descriptions, utilizing the pre-trained encoder to initialize the CLSP encoder. Then, we deploy contrastive learning to train the CLSP encoder to effectively represent precise state information. Additionally, we enhance the representation of numerical information using the Random Fourier Features (RFF) method for high-fidelity mapping. Extensive experiments demonstrate the superior precision and generalization capabilities of our representation, achieving outstanding results in text-state retrieval, reinforcement learning navigation tasks, and multimodal large language model understanding.

cs.AI↗

arXiv2024

PromptTA: Prompt-driven Text Adapter for Source-free Domain Generalization

Source-free domain generalization (SFDG) tackles the challenge of adapting models to unseen target domains without access to source domain data. To deal with this challenging task, recent advances in SFDG have primarily focused on leveraging the text modality of vision-language models such as CLIP. These methods involve developing a transferable linear classifier based on diverse style features extracted from the text and learned prompts or deriving domain-unified text representations from domain banks. However, both style features and domain banks have limitations in capturing comprehensive domain knowledge. In this work, we propose Prompt-Driven Text Adapter (PromptTA) method, which is designed to better capture the distribution of style features and employ resampling to ensure thorough coverage of domain knowledge. To further leverage this rich domain information, we introduce a text adapter that learns from these style features for efficient domain information storage. Extensive experiments conducted on four benchmark datasets demonstrate that PromptTA achieves state-of-the-art performance. The code is available at https://github.com/zhanghr2001/PromptTA.

cs.CV↗

arXiv2024

Single-View 3D Reconstruction via SO(2)-Equivariant Gaussian Sculpting Networks

This paper introduces SO(2)-Equivariant Gaussian Sculpting Networks (GSNs) as an approach for SO(2)-Equivariant 3D object reconstruction from single-view image observations. GSNs take a single observation as input to generate a Gaussian splat representation describing the observed object's geometry and texture. By using a shared feature extractor before decoding Gaussian colors, covariances, positions, and opacities, GSNs achieve extremely high throughput (>150FPS). Experiments demonstrate that GSNs can be trained efficiently using a multi-view rendering loss and are competitive, in quality, with expensive diffusion-based reconstruction algorithms. The GSN model is validated on multiple benchmark experiments. Moreover, we demonstrate the potential for GSNs to be used within a robotic manipulation pipeline for object-centric grasping.

cs.CV↗

arXiv2024

Computer-generated holography enables high-uniformity, high-efficiency depth-of-focus extension in endoscopic OCT

Fiber-form optics extends the high-resolution tomographic imaging capabilities of Optical Coherence Tomography (OCT) to the inside of the human body, i.e., endoscopic OCT. However, it still faces challenges due to the trade-off between probe size, resolution, and Depth Of Focus (DOF). Here we introduce a method for extending the DOF in endoscopic OCT with high uniformity and efficiency. On the basis of multi-level diffractive optics, we leverage the multi-dimensional light field modulation capabilities of Computer-Generated Holography (CGH), to achieve precise control of the intensity distribution of the off-axis portion of the OCT probe light. Our method eliminates the need for an objective lens, allowing for direct fabrication at the distal facet of a single-mode fiber using femtosecond laser two-photon 3D printing. The superiority of our method has been verified through numerical simulation, beam measurement, and imaging results obtained with our home-built endoscopic OCT system.

physics.optics↗

arXiv2024

Curriculum Prompting Foundation Models for Medical Image Segmentation

Adapting large pre-trained foundation models, e.g., SAM, for medical image segmentation remains a significant challenge. A crucial step involves the formulation of a series of specialized prompts that incorporate specific clinical instructions. Past works have been heavily reliant on a singular type of prompt for each instance, necessitating manual input of an ideally correct prompt, which is less efficient. To tackle this issue, we propose to utilize prompts of different granularity, which are sourced from original images to provide a broader scope of clinical insights. However, combining prompts of varying types can pose a challenge due to potential conflicts. In response, we have designed a coarse-to-fine mechanism, referred to as curriculum prompting, that progressively integrates prompts of different types. Through extensive experiments on three public medical datasets across various modalities, we demonstrate the effectiveness of our proposed approach, which not only automates the prompt generation process but also yields superior performance compared to other SAM-based medical image segmentation methods. Code is available at: https://github.com/AnnaZzz-zxq/Curriculum-Prompting.

cs.CV↗

arXiv2024

Room-temperature Optically Detected Magnetic Resonance of Telecom Single Photon Emitters in GaN

Solid-state defects susceptible of spin manipulation hold great promise for scalable quantum technology. To broaden their utility, operating at room temperature and emitting in the telecom wavelength range are desired, eliminating cryogenic requirements and leveraging existing optical fiber infrastructure for transmitting the quantum information. To that end, we report that telecom single photon emitters (SPEs) in gallium nitride (GaN) exhibit optically detected magnetic resonance (ODMR) at room temperature. The analysis of ODMR as a function of magnetic field orientation enables the determination of the orientation of the spin quantization axis with respect to the GaN crystalline lattice. The optical transitions dynamics are analyzed to gain further insight into the transition rates dominating ODMR. Our findings, coupled with GaN's mature fabrication technology, could facilitate the realization of scalable quantum technology.

cond-mat.mtrl-sci↗

arXiv2024

NP-TCMtarget: a network pharmacology platform for exploring mechanisms of action of Traditional Chinese medicine

The biological targets of traditional Chinese medicine (TCM) are the core effectors mediating the interaction between TCM and the human body. Identification of TCM targets is essential to elucidate the chemical basis and mechanisms of TCM for treating diseases. Given the chemical complexity of TCM, both in silico high-throughput drug-target interaction predicting models and biological profile-based methods have been commonly applied for identifying TCM targets based on the structural information of TCM chemical components and biological information, respectively. However, the existing methods lack the integration of TCM chemical and biological information, resulting in difficulty in the systematic discovery of TCM action pathways. To solve this problem, we propose a novel target identification model NP-TCMtarget to explore the TCM target path by combining the overall chemical and biological profiles. First, NP-TCMtarget infers TCM effect targets by calculating associations between drug/disease inducible gene expression profiles and specific gene signatures for 8,233 targets. Then, NP-TCMtarget utilizes a constructed binary classification model to predict binding targets of herbal ingredients. Finally, we can distinguish TCM direct and indirect targets by comparing the effect targets and binding targets to establish the action pathways of herbal components-direct targets-indirect targets by mapping TCM targets in the biological molecular network. We apply NP-TCMtarget to the formula XiaoKeAn to demonstrate the power of revealing the action pathways of herbal formula. We expect that this novel model could provide a systematic framework for exploring the molecular mechanisms of TCM at the target level. NP-TCMtarget is available at http://www.bcxnfz.top/NP-TCMtarget.

q-bio.MN↗