Chao Wang

Yu-Chen Lin, Akhilesh Kumar, Norman Chang et al.

We present four main contributions to enhance the performance of Large Language Models (LLMs) in generating domain-specific code: (i) utilizing LLM-based data splitting and data renovation techniques to improve the semantic representation of embeddings' space; (ii) introducing the Chain of Density for Renovation Credibility (CoDRC), driven by LLMs, and the Adaptive Text Renovation (ATR) algorithm for assessing data renovation reliability; (iii) developing the Implicit Knowledge Expansion and Contemplation (IKEC) Prompt technique; and (iv) effectively refactoring existing scripts to generate new and high-quality scripts with LLMs. By using engineering simulation software RedHawk-SC as a case study, we demonstrate the effectiveness of our data pre-processing method for expanding and categorizing scripts. When combined with IKEC, these techniques enhance the Retrieval-Augmented Generation (RAG) method in retrieving more relevant information, ultimately achieving a 73.33% "Percentage of Correct Lines" for code generation problems in MapReduce applications.

Zhongjiang He, Zihan Wang, Xinzhang Liu et al.

In this technical report, we present TeleChat, a collection of large language models (LLMs) with parameters of 3 billion, 7 billion and 12 billion. It includes pretrained language models as well as fine-tuned chat models that is aligned with human preferences. TeleChat is initially pretrained on an extensive corpus containing a diverse collection of texts from both English and Chinese languages, including trillions of tokens. Subsequently, the model undergoes fine-tuning to align with human preferences, following a detailed methodology that we describe. We evaluate the performance of TeleChat on various tasks, including language understanding, mathematics, reasoning, code generation, and knowledge-based question answering. Our findings indicate that TeleChat achieves comparable performance to other open-source models of similar size across a wide range of public benchmarks. To support future research and applications utilizing LLMs, we release the fine-tuned model checkpoints of TeleChat's 7B and 12B variant, along with code and a portion of our pretraining data, to the public community.

Shaohua Wu, Xudong Zhao, Shenling Wang et al.

In this work, we develop and release Yuan 2.0, a series of large language models with parameters ranging from 2.1 billion to 102.6 billion. The Localized Filtering-based Attention (LFA) is introduced to incorporate prior knowledge of local dependencies of natural language into Attention. A data filtering and generating system is presented to build pre-training and fine-tuning dataset in high quality. A distributed training method with non-uniform pipeline parallel, data parallel, and optimizer parallel is proposed, which greatly reduces the bandwidth requirements of intra-node communication, and achieves good performance in large-scale distributed training. Yuan 2.0 models display impressive ability in code generation, math problem-solving, and chatting compared with existing models. The latest version of YUAN 2.0, including model weights and source code, is accessible at Github.

Chenxu Zhang, Chao Wang, Jianfeng Zhang et al.

The generation of emotional talking faces from a single portrait image remains a significant challenge. The simultaneous achievement of expressive emotional talking and accurate lip-sync is particularly difficult, as expressiveness is often compromised for the accuracy of lip-sync. As widely adopted by many prior works, the LSTM network often fails to capture the subtleties and variations of emotional expressions. To address these challenges, we introduce DREAM-Talk, a two-stage diffusion-based audio-driven framework, tailored for generating diverse expressions and accurate lip-sync concurrently. In the first stage, we propose EmoDiff, a novel diffusion module that generates diverse highly dynamic emotional expressions and head poses in accordance with the audio and the referenced emotion style. Given the strong correlation between lip motion and audio, we then refine the dynamics with enhanced lip-sync accuracy using audio features and emotion style. To this end, we deploy a video-to-video rendering module to transfer the expressions and lip motions from our proxy 3D avatar to an arbitrary portrait. Both quantitatively and qualitatively, DREAM-Talk outperforms state-of-the-art methods in terms of expressiveness, lip-sync accuracy and perceptual quality.

Chia-Hao Chiang, Zheng-Han Huang, Liwen Liu et al.

Human activities accelerate consumption of fossil fuels and produce greenhouse gases, resulting in urgent issues today: global warming and the climate change. These indirectly cause severe natural disasters, plenty of lives suffering and huge losses of agricultural properties. To mitigate impacts on our lands, scientists are developing renewable, reusable, and clean energies and climatologists are trying to predict the extremes. Meanwhile, governments are publicizing resource-saving policies for a more eco-friendly society and arousing environment awareness. One of the most influencing factors is the precipitation, bringing condensed water vapor onto lands. Water resources are the most significant but basic needs in society, not only supporting our livings, but also economics. In Taiwan, although the average annual precipitation is up to 2,500 millimeter (mm), the water allocation for each person is lower than the global average due to drastically geographical elevation changes and uneven distribution through the year. Thus, it is crucial to track and predict the rainfall to make the most use of it and to prevent the floods. However, climate models have limited resolution and require intensive computational power for local-scale use. Therefore, we proposed a deep convolutional neural network with skip connections, attention blocks, and auxiliary data concatenation, in order to downscale the low-resolution precipitation data into high-resolution one. Eventually, we compare with other climate downscaling methods and show better performance in metrics of Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Pearson Correlation, structural similarity index (SSIM), and forecast indicators.

Chao Wang, Zhihao Xia, Thomas Leimkuehler et al.

While consumer displays increasingly support more than 10 stops of dynamic range, most image assets such as internet photographs and generative AI content remain limited to 8-bit low dynamic range (LDR), constraining their utility across high dynamic range (HDR) applications. Currently, no generative model can produce high-bit, high-dynamic range content in a generalizable way. Existing LDR-to-HDR conversion methods often struggle to produce photorealistic details and physically-plausible dynamic range in the clipped areas. We introduce LEDiff, a method that enables a generative model with HDR content generation through latent space fusion inspired by image-space exposure fusion techniques. It also functions as an LDR-to-HDR converter, expanding the dynamic range of existing low-dynamic range images. Our approach uses a small HDR dataset to enable a pretrained diffusion model to recover detail and dynamic range in clipped highlights and shadows. LEDiff brings HDR capabilities to existing generative models and converts any LDR image to HDR, creating photorealistic HDR outputs for image generation, image-based lighting (HDR environment map generation), and photographic effects such as depth of field simulation, where linear HDR data is essential for realistic quality.

Chao Wang, Weiwei Fu, Yang Zhou

Vision-language models (VLMs) have achieved remarkable advancements, capitalizing on the impressive capabilities of large language models (LLMs) across diverse tasks. Despite this, a critical challenge known as hallucination occurs when models overconfidently describe objects or attributes absent from the image, a problem exacerbated by the tendency of VLMs to rely on linguistic priors. This limitation reduces model reliability in high-stakes applications. In this work, we have observed the characteristic of logits' continuity consistency enhancement and introduced a straightforward and efficient method, Cross-Temporal Prediction Connection (TPC), designed to enhance the semantic consistency of logits by connecting them temporally across timesteps. TPC amplifies information flow and improves coherence, effectively reducing hallucination. Extensive experiments show that TPC surpasses existing representatives, delivering superior performance in both accuracy and efficiency while maintaining robustness in open-ended text generation tasks.

Mincong Huang, Chao Wang, Chi Ma et al.

Pipeline parallelism is an essential technique in the training of large-scale Transformer models. However, it suffers from imbalanced memory consumption, leading to insufficient memory utilization. The BPipe technique was proposed to address this issue and has proven effective in the GPT-3 model. Nevertheless, our experiments have not yielded similar benefits for LLaMA training. Additionally, BPipe only yields negligible benefits for GPT-3 training when applying flash attention. We analyze the underlying causes of the divergent performance of BPipe on GPT-3 and LLaMA. Furthermore, we introduce a novel method to estimate the performance of BPipe.

Qianli Zhao, Chao Wang, Richard Gerlach et al.

Realised volatility has become increasingly prominent in volatility forecasting due to its ability to capture intraday price fluctuations. With a growing variety of realised volatility estimators, each with unique advantages and limitations, selecting an optimal estimator may introduce challenges. In this thesis, aiming to synthesise the impact of various realised volatility measures on volatility forecasting, we propose an extension of the Realised GARCH model that incorporates an autoencoder-generated synthetic realised measure, combining the information from multiple realised measures in a nonlinear manner. Our proposed model extends existing linear methods, such as Principal Component Analysis and Independent Component Analysis, to reduce the dimensionality of realised measures. The empirical evaluation, conducted across four major stock markets from January 2000 to June 2022 and including the period of COVID-19, demonstrates both the feasibility of applying an autoencoder to synthesise volatility measures and the superior effectiveness of the proposed model in one-step-ahead rolling volatility forecasting. The model exhibits enhanced flexibility in parameter estimations across each rolling window, outperforming traditional linear approaches. These findings indicate that nonlinear dimension reduction offers further adaptability and flexibility in improving the synthetic realised measure, with promising implications for future volatility forecasting applications.

Ying Zhang, Peng Zhang, Mincong Huang et al.

Quantization is a proven effective method for compressing large language models. Although popular techniques like W8A8 and W4A16 effectively maintain model performance, they often fail to concurrently speed up the prefill and decoding stages of inference. W4A8 is a promising strategy to accelerate both of them while usually leads to a significant performance degradation. To address these issues, we present QQQ, a Quality Quattuor-bit Quantization method with 4-bit weights and 8-bit activations. QQQ employs adaptive smoothing and Hessian-based compensation, significantly enhancing the performance of quantized models without extensive training. Furthermore, we meticulously engineer W4A8 GEMM kernels to increase inference speed. Our specialized per-channel W4A8 GEMM and per-group W4A8 GEMM achieve impressive speed increases of 3.67$\times$ and 3.29 $\times$ over FP16 GEMM. Our extensive experiments show that QQQ achieves performance on par with existing state-of-the-art LLM quantization methods while significantly accelerating inference, achieving speed boosts up to 2.24 $\times$, 2.10$\times$, and 1.25$\times$ compared to FP16, W8A8, and W4A16, respectively.