arXiv subjects

Jiajun Zhang

Publications and source records attributed to Jiajun Zhang.

At least 145 records · Page 8Linked to original sources

arXiv2024

BLSP-Emo: Towards Empathetic Large Speech-Language Models

The recent release of GPT-4o showcased the potential of end-to-end multimodal models, not just in terms of low latency but also in their ability to understand and generate expressive speech with rich emotions. While the details are unknown to the open research community, it likely involves significant amounts of curated data and compute, neither of which is readily accessible. In this paper, we present BLSP-Emo (Bootstrapped Language-Speech Pretraining with Emotion support), a novel approach to developing an end-to-end speech-language model capable of understanding both semantics and emotions in speech and generate empathetic responses. BLSP-Emo utilizes existing speech recognition (ASR) and speech emotion recognition (SER) datasets through a two-stage process. The first stage focuses on semantic alignment, following recent work on pretraining speech-language models using ASR data. The second stage performs emotion alignment with the pretrained speech-language model on an emotion-aware continuation task constructed from SER data. Our experiments demonstrate that the BLSP-Emo model excels in comprehending speech and delivering empathetic responses, both in instruction-following tasks and conversations.

cs.CL↗

arXiv2024

SpikeLM: Towards General Spike-Driven Language Modeling via Elastic Bi-Spiking Mechanisms

Towards energy-efficient artificial intelligence similar to the human brain, the bio-inspired spiking neural networks (SNNs) have advantages of biological plausibility, event-driven sparsity, and binary activation. Recently, large-scale language models exhibit promising generalization capability, making it a valuable issue to explore more general spike-driven models. However, the binary spikes in existing SNNs fail to encode adequate semantic information, placing technological challenges for generalization. This work proposes the first fully spiking mechanism for general language tasks, including both discriminative and generative ones. Different from previous spikes with {0,1} levels, we propose a more general spike formulation with bi-directional, elastic amplitude, and elastic frequency encoding, while still maintaining the addition nature of SNNs. In a single time step, the spike is enhanced by direction and amplitude information; in spike frequency, a strategy to control spike firing rate is well designed. We plug this elastic bi-spiking mechanism in language modeling, named SpikeLM. It is the first time to handle general language tasks with fully spike-driven models, which achieve much higher accuracy than previously possible. SpikeLM also greatly bridges the performance gap between SNNs and ANNs in language modeling. Our code is available at https://github.com/Xingrun-Xing/SpikeLM.

cs.NE↗

arXiv2024

CIF-Bench: A Chinese Instruction-Following Benchmark for Evaluating the Generalizability of Large Language Models

The advancement of large language models (LLMs) has enhanced the ability to generalize across a wide range of unseen natural language processing (NLP) tasks through instruction-following. Yet, their effectiveness often diminishes in low-resource languages like Chinese, exacerbated by biased evaluations from data leakage, casting doubt on their true generalizability to new linguistic territories. In response, we introduce the Chinese Instruction-Following Benchmark (CIF-Bench), designed to evaluate the zero-shot generalizability of LLMs to the Chinese language. CIF-Bench comprises 150 tasks and 15,000 input-output pairs, developed by native speakers to test complex reasoning and Chinese cultural nuances across 20 categories. To mitigate data contamination, we release only half of the dataset publicly, with the remainder kept private, and introduce diversified instructions to minimize score variance, totaling 45,000 data instances. Our evaluation of 28 selected LLMs reveals a noticeable performance gap, with the best model scoring only 52.9%, highlighting the limitations of LLMs in less familiar language and task contexts. This work not only uncovers the current limitations of LLMs in handling Chinese language tasks but also sets a new standard for future LLM generalizability research, pushing towards the development of more adaptable, culturally informed, and linguistically diverse models.

cs.CL↗

arXiv2024

Diver: Large Language Model Decoding with Span-Level Mutual Information Verification

Large language models (LLMs) have shown impressive capabilities in adapting to various tasks when provided with task-specific instructions. However, LLMs using standard decoding strategies often struggle with deviations from the inputs. Intuitively, compliant LLM outputs should reflect the information present in the input, which can be measured by point-wise mutual information (PMI) scores. Therefore, we propose Diver, a novel approach that enhances LLM Decoding through span-level PMI verification. During inference, Diver first identifies divergence steps that may lead to multiple candidate spans. Subsequently, it calculates the PMI scores by assessing the log-likelihood gains of the input if the candidate spans are generated. Finally, the optimal span is selected based on the PMI re-ranked output distributions. We evaluate our method across various downstream tasks, and empirical results demonstrate that Diver significantly outperforms existing decoding methods in both performance and versatility.

cs.CL↗

arXiv2024

X-Instruction: Aligning Language Model in Low-resource Languages with Self-curated Cross-lingual Instructions

Large language models respond well in high-resource languages like English but struggle in low-resource languages. It may arise from the lack of high-quality instruction following data in these languages. Directly translating English samples into these languages can be a solution but unreliable, leading to responses with translation errors and lacking language-specific or cultural knowledge. To address this issue, we propose a novel method to construct cross-lingual instruction following samples with instruction in English and response in low-resource languages. Specifically, the language model first learns to generate appropriate English instructions according to the natural web texts in other languages as responses. The candidate cross-lingual instruction tuning samples are further refined and diversified. We have employed this method to build a large-scale cross-lingual instruction tuning dataset on 10 languages, namely X-Instruction. The instruction data built using our method incorporate more language-specific knowledge compared with the naive translation method. Experimental results have shown that the response quality of the model tuned on X-Instruction greatly exceeds the model distilled from a powerful teacher model, reaching or even surpassing the ones of ChatGPT. In addition, we find that models tuned on cross-lingual instruction following samples can follow the instruction in the output language without further tuning.

cs.CL↗

arXiv2024

BLSP-KD: Bootstrapping Language-Speech Pre-training via Knowledge Distillation

Recent end-to-end approaches have shown promise in extending large language models (LLMs) to speech inputs, but face limitations in directly assessing and optimizing alignment quality and fail to achieve fine-grained alignment due to speech-text length mismatch. We introduce BLSP-KD, a novel approach for Bootstrapping Language-Speech Pretraining via Knowledge Distillation, which addresses these limitations through two key techniques. First, it optimizes speech-text alignment by minimizing the divergence between the LLM's next-token prediction distributions for speech and text inputs using knowledge distillation. Second, it employs a continuous-integrate-andfire strategy to segment speech into tokens that correspond one-to-one with text tokens, enabling fine-grained alignment. We also introduce Partial LoRA (PLoRA), a new adaptation method supporting LLM finetuning for speech inputs under knowledge distillation. Quantitative evaluation shows that BLSP-KD outperforms previous end-to-end baselines and cascaded systems with comparable scale of parameters, facilitating general instruction-following capabilities for LLMs with speech inputs. This approach provides new possibilities for extending LLMs to spoken language interactions.

cs.CL↗

arXiv2024

BLSP: Bootstrapping Language-Speech Pre-training via Behavior Alignment of Continuation Writing

The emergence of large language models (LLMs) has sparked significant interest in extending their remarkable language capabilities to speech. However, modality alignment between speech and text still remains an open problem. Current solutions can be categorized into two strategies. One is a cascaded approach where outputs (tokens or states) of a separately trained speech recognition system are used as inputs for LLMs, which limits their potential in modeling alignment between speech and text. The other is an end-to-end approach that relies on speech instruction data, which is very difficult to collect in large quantities. In this paper, we address these issues and propose the BLSP approach that Bootstraps Language-Speech Pre-training via behavior alignment of continuation writing. We achieve this by learning a lightweight modality adapter between a frozen speech encoder and an LLM, ensuring that the LLM exhibits the same generation behavior regardless of the modality of input: a speech segment or its transcript. The training process can be divided into two steps. The first step prompts an LLM to generate texts with speech transcripts as prefixes, obtaining text continuations. In the second step, these continuations are used as supervised signals to train the modality adapter in an end-to-end manner. We demonstrate that this straightforward process can extend the capabilities of LLMs to speech, enabling speech recognition, speech translation, spoken language understanding, and speech conversation, even in zero-shot cross-lingual scenarios.

cs.CL↗

arXiv2024

New galaxy UV luminosity constraints on warm dark matter from JWST

We exploit the recent {\it James Webb Space Telescope} (JWST) determination of galaxy UV luminosity functions over the redshift range $z=9-14.5$ to derive constraints on warm dark matter (WDM) models. The delayed structure formation in WDM universes make high-redshift observations a powerful probe to set limits on the particle mass $m_\mathrm{x}$ of WDM candidates. By integrating these observations with blank-field surveys conducted by the {\it Hubble Space Telescope} (HST) at redshifts $z=4-8$, we impose constraints on both astrophysical parameters ($β$, $γ$, $ε_{\mathrm N}$, $M_c$ for a double-power law star formation efficiency, and $σ_{M_{\mathrm{UV}}}$ for a Gaussian magnitude-halo mass relation) and the WDM parameter (dark matter particle mass $m_\mathrm{x}$) simultaneously. We find a new limit of $m_\mathrm{x} \geq 3.2$ keV for the mass of thermal relic WDM particles at $95\%$ confidence level. This bound is tighter than the most stringent result derived using HST data before. Future JWST observations could further reduce the observation uncertainties and improve this constraint.

astro-ph.CO↗

arXiv2024

Bridging the Gap between Different Vocabularies for LLM Ensemble

Ensembling different large language models (LLMs) to unleash their complementary potential and harness their individual strengths is highly valuable. Nevertheless, vocabulary discrepancies among various LLMs have constrained previous studies to either selecting or blending completely generated outputs. This limitation hinders the dynamic correction and enhancement of outputs during the generation process, resulting in a limited capacity for effective ensemble. To address this issue, we propose a novel method to Ensemble LLMs via Vocabulary Alignment (EVA). EVA bridges the lexical gap among various LLMs, enabling meticulous ensemble at each generation step. Specifically, we first learn mappings between the vocabularies of different LLMs with the assistance of overlapping tokens. Subsequently, these mappings are employed to project output distributions of LLMs into a unified space, facilitating a fine-grained ensemble. Finally, we design a filtering strategy to exclude models that generate unfaithful tokens. Experimental results on commonsense reasoning, arithmetic reasoning, machine translation, and data-to-text generation tasks demonstrate the superiority of our approach compared with individual LLMs and previous ensemble methods conducted on complete outputs. Further analyses confirm that our approach can leverage knowledge from different language models and yield consistent improvement.

cs.CL↗

arXiv2024

Ins-HOI: Instance Aware Human-Object Interactions Recovery

Accurately modeling detailed interactions between human/hand and object is an appealing yet challenging task. Current multi-view capture systems are only capable of reconstructing multiple subjects into a single, unified mesh, which fails to model the states of each instance individually during interactions. To address this, previous methods use template-based representations to track human/hand and object. However, the quality of the reconstructions is limited by the descriptive capabilities of the templates so that these methods are inherently struggle with geometry details, pressing deformations and invisible contact surfaces. In this work, we propose an end-to-end Instance-aware Human-Object Interactions recovery (Ins-HOI) framework by introducing an instance-level occupancy field representation. However, the real-captured data is presented as a holistic mesh, unable to provide instance-level supervision. To address this, we further propose a complementary training strategy that leverages synthetic data to introduce instance-level shape priors, enabling the disentanglement of occupancy fields for different instances. Specifically, synthetic data, created by randomly combining individual scans of humans/hands and objects, guides the network to learn a coarse prior of instances. Meanwhile, real-captured data helps in learning the overall geometry and restricting interpenetration in contact areas. As demonstrated in experiments, our method Ins-HOI supports instance-level reconstruction and provides reasonable and realistic invisible contact surfaces even in cases of extremely close interaction. To facilitate the research of this task, we collect a large-scale, high-fidelity 3D scan dataset, including 5.2k high-quality scans with real-world human-chair and hand-object interactions. The code and data will be public for research purposes.

cs.CV↗

arXiv2024

21-cm foreground removal using AI and frequency-difference technique

The deep learning technique has been employed in removing foreground contaminants from 21 cm intensity mapping, but its effectiveness is limited by the large dynamic range of the foreground amplitude. In this study, we develop a novel foreground removal technique grounded in U-Net networks. The essence of this technique lies in introducing an innovative data preprocessing step specifically, utilizing the temperature difference between neighboring frequency bands as input, which can substantially reduce the dynamic range of foreground amplitudes by approximately two orders of magnitude. This reduction proves to be highly advantageous for the U-Net foreground removal. We observe that the HI signal can be reliably recovered, as indicated by the cross-correlation power spectra showing unity agreement at the scale of $k < 0.3 h^{-1}$Mpc in the absence of instrumental effects. Moreover, accounting for the systematic beam effects, our reconstruction displays consistent auto-correlation and cross-correlation power spectrum ratios at the $1σ$ level across scales $k \lesssim 0.1 h^{-1}$Mpc, with only a 10% reduction observed in the cross-correlation power spectrum at $k\simeq0.2 h^{-1}$Mpc. The effects of redshift-space distortion are also reconstructed successfully, as evidenced by the quadrupole power spectra matching. In comparison, our method outperforms the traditional Principal Component Analysis method, which derived cross-correlation ratios are underestimated by around 60%. We simulated various white noise levels in the map and found that the mean cross-correlation ratio $\bar{R}_\mathrm{cross} \gtrsim 0.8$ when the level of the thermal noise is smaller than or equal to that of the HI signal. We conclude that the proposed frequency-difference technique can significantly enhance network performance by reducing the amplitude range of foregrounds and aiding in the prevention of HI loss.

astro-ph.CO↗

arXiv2024

DPPA: Pruning Method for Large Language Model to Model Merging

Model merging is to combine fine-tuned models derived from multiple domains, with the intent of enhancing the model's proficiency across various domains. The principal concern is the resolution of parameter conflicts. A substantial amount of existing research remedy this issue during the merging stage, with the latest study focusing on resolving this issue throughout the pruning stage. The DARE approach has exhibited promising outcomes when applied to a simplistic fine-tuned model. However, the efficacy of this method tends to wane when employed on complex fine-tuned models that show a significant parameter bias relative to the baseline model. In this paper, we introduce a dual-stage method termed Dynamic Pruning Partition Amplification (DPPA), devised to tackle the challenge of merging complex fine-tuned models. Initially, we introduce Dynamically Pruning (DP), an improved approach based on magnitude pruning, which aim is to enhance performance at higher pruning rates. Subsequently, we propose Dynamically Partition Amplification (DPA), a rescaling strategy, is designed to dynamically amplify parameter partitions in relation to their significance levels. The experimental results show that our method maintains a mere 20% of domain-specific parameters and yet delivers a performance comparable to other methodologies that preserve up to 90% of parameters. Furthermore, our method displays outstanding performance post-pruning, leading to a significant improvement of nearly 20% performance in model merging. We make our code on Github.

cs.CL↗

arXiv2023

ProxyCap: Real-time Monocular Full-body Capture in World Space via Human-Centric Proxy-to-Motion Learning

Learning-based approaches to monocular motion capture have recently shown promising results by learning to regress in a data-driven manner. However, due to the challenges in data collection and network designs, it remains challenging for existing solutions to achieve real-time full-body capture while being accurate in world space. In this work, we introduce ProxyCap, a human-centric proxy-to-motion learning scheme to learn world-space motions from a proxy dataset of 2D skeleton sequences and 3D rotational motions. Such proxy data enables us to build a learning-based network with accurate world-space supervision while also mitigating the generalization issues. For more accurate and physically plausible predictions in world space, our network is designed to learn human motions from a human-centric perspective, which enables the understanding of the same motion captured with different camera trajectories. Moreover, a contact-aware neural motion descent module is proposed in our network so that it can be aware of foot-ground contact and motion misalignment with the proxy observations. With the proposed learning-based solution, we demonstrate the first real-time monocular full-body capture system with plausible foot-ground contact in world space even using hand-held moving cameras. Our project page is https://zhangyux15.github.io/ProxyCapV2.

cs.CV↗

arXiv2023

Toward Real World Stereo Image Super-Resolution via Hybrid Degradation Model and Discriminator for Implied Stereo Image Information

Real-world stereo image super-resolution has a significant influence on enhancing the performance of computer vision systems. Although existing methods for single-image super-resolution can be applied to improve stereo images, these methods often introduce notable modifications to the inherent disparity, resulting in a loss in the consistency of disparity between the original and the enhanced stereo images. To overcome this limitation, this paper proposes a novel approach that integrates a implicit stereo information discriminator and a hybrid degradation model. This combination ensures effective enhancement while preserving disparity consistency. The proposed method bridges the gap between the complex degradations in real-world stereo domain and the simpler degradations in real-world single-image super-resolution domain. Our results demonstrate impressive performance on synthetic and real datasets, enhancing visual perception while maintaining disparity consistency. The complete code is available at the following \href{https://github.com/fzuzyb/SCGLANet}{link}.

eess.IV↗

arXiv2023

MoDS: Model-oriented Data Selection for Instruction Tuning

Instruction tuning has become the de facto method to equip large language models (LLMs) with the ability of following user instructions. Usually, hundreds of thousands or millions of instruction-following pairs are employed to fine-tune the foundation LLMs. Recently, some studies show that a small number of high-quality instruction data is enough. However, how to select appropriate instruction data for a given LLM is still an open problem. To address this problem, in this paper we present a model-oriented data selection (MoDS) approach, which selects instruction data based on a new criteria considering three aspects: quality, coverage and necessity. First, our approach utilizes a quality evaluation model to filter out the high-quality subset from the original instruction dataset, and then designs an algorithm to further select from the high-quality subset a seed instruction dataset with good coverage. The seed dataset is applied to fine-tune the foundation LLM to obtain an initial instruction-following LLM. Finally, we develop a necessity evaluation model to find out the instruction data which are performed badly in the initial instruction-following LLM and consider them necessary instructions to further improve the LLMs. In this way, we can get a small high-quality, broad-coverage and high-necessity subset from the original instruction datasets. Experimental results show that, the model fine-tuned with 4,000 instruction pairs selected by our approach could perform better than the model fine-tuned with the full original dataset which includes 214k instruction data.

cs.CL↗

arXiv2023

BigTranslate: Augmenting Large Language Models with Multilingual Translation Capability over 100 Languages

Large language models (LLMs) demonstrate promising translation performance among various natural languages. However, many LLMs especially the open-sourced ones, such as BLOOM and LLaMA, are English-dominant and support only dozens of natural languages, making the potential of LLMs on language translation less explored. In this work, we present BigTranslate which adapts LLaMA that covers only 20 languages and enhances it with multilingual translation capability on more than 100 languages. BigTranslate is built upon LLaMA-13B and it is optimized in three steps. First, we continue training LLaMA with massive Chinese monolingual data. Second, we continue training the model with a large-scale parallel dataset that covers 102 natural languages. Third, we instruct-tune the foundation model with multilingual translation instructions, leading to our BigTranslate model. The preliminary experiments on multilingual translation show that BigTranslate performs comparably with ChatGPT and Google Translate in many languages and even outperforms ChatGPT in 8 language pairs. We release the BigTranslate model and hope it can advance the research progress.

cs.CL↗

arXiv2023

ChineseWebText: Large-scale High-quality Chinese Web Text Extracted with Effective Evaluation Model

During the development of large language models (LLMs), the scale and quality of the pre-training data play a crucial role in shaping LLMs' capabilities. To accelerate the research of LLMs, several large-scale datasets, such as C4 [1], Pile [2], RefinedWeb [3] and WanJuan [4], have been released to the public. However, most of the released corpus focus mainly on English, and there is still lack of complete tool-chain for extracting clean texts from web data. Furthermore, fine-grained information of the corpus, e.g. the quality of each text, is missing. To address these challenges, we propose in this paper a new complete tool-chain EvalWeb to extract Chinese clean texts from noisy web data. First, similar to previous work, manually crafted rules are employed to discard explicit noisy texts from the raw crawled web contents. Second, a well-designed evaluation model is leveraged to assess the remaining relatively clean data, and each text is assigned a specific quality score. Finally, we can easily utilize an appropriate threshold to select the high-quality pre-training data for Chinese. Using our proposed approach, we release the largest and latest large-scale high-quality Chinese web text ChineseWebText, which consists of 1.42 TB and each text is associated with a quality score, facilitating the LLM researchers to choose the data according to the desired quality thresholds. We also release a much cleaner subset of 600 GB Chinese data with the quality exceeding 90%.

cs.CL↗

arXiv2023

The BINGO Project IX: Search for Fast Radio Bursts -- A Forecast for the BINGO Interferometry System

The Baryon Acoustic Oscillations (BAO) from Integrated Neutral Gas Observations (BINGO) radio telescope will use the neutral Hydrogen emission line to map the Universe in the redshift range $0.127 \le z \le 0.449$, with the main goal of probing BAO. In addition, the instrument optical design and hardware configuration support the search for Fast Radio Bursts (FRBs). In this work, we propose the use of a BINGO Interferometry System (BIS) including new auxiliary, smaller, radio telescopes (hereafter \emph{outriggers}). The interferometric approach makes it possible to pinpoint the FRB sources in the sky. We present here the results of several BIS configurations combining BINGO horns with and without mirrors ($4$ m, $5$ m, and $6$ m) and 5, 7, 9, or 10 for single horns. We developed a new {\tt Python} package, the {\tt FRBlip}, which generates synthetic FRB mock catalogs and computes, based on a telescope model, the observed signal-to-noise ratio (S/N) that we used to compute numerically the detection rates of the telescopes and how many interferometry pairs of telescopes (\emph{baselines}) can observe an FRB. FRBs observed by more than one baseline are the ones whose location can be determined. We thus evaluate the performance of BIS regarding FRB localization. We found that BIS will be able to localize 23 FRBs yearly with single horn outriggers in the best configuration (using 10 outriggers of 6 m mirrors), with redshift $z \leq 0.96$; the full localization capability depends on the number and the type of the outriggers. Wider beams are best to pinpoint FRB sources because potential candidates will be observed by more baselines, while narrow beams look deep in redshift. The BIS can be a powerful extension of the regular BINGO telescope, dedicated to observe hundreds of FRBs during Phase 1. Many of them will be well localized with a single horn + 6 m dish as outriggers.(Abridged)

astro-ph.IM↗