In this paper, we introduce a novel dynamic expert selection framework for Mixture of Experts (MoE) models, aiming to enhance computational efficiency and model performance by adjusting the number of activated experts based on input difficulty. Unlike traditional MoE approaches that rely on fixed Top-K routing, which activates a predetermined number of experts regardless of the input's complexity, our method dynamically selects experts based on the confidence level in expert selection for each input. This allows for a more efficient utilization of computational resources, activating more experts for complex tasks requiring advanced reasoning and fewer for simpler tasks. Through extensive evaluations, our dynamic routing method demonstrates substantial improvements over conventional Top-2 routing across various benchmarks, achieving an average improvement of 0.7% with less than 90% activated parameters. Further analysis shows our model dispatches more experts to tasks requiring complex reasoning skills, like BBH, confirming its ability to dynamically allocate computational resources in alignment with the input's complexity. Our findings also highlight a variation in the number of experts needed across different layers of the transformer model, offering insights into the potential for designing heterogeneous MoE frameworks. The code and models are available at https://github.com/ZhenweiAn/Dynamic_MoE.
The success of large language models has inspired researchers to transfer their exceptional representing ability to other modalities. Several recent works leverage image-caption alignment datasets to train multimodal large language models (MLLMs), which achieve state-of-the-art performance on image-to-text tasks. However, there are very few studies exploring whether MLLMs truly understand the complete image information, i.e., global information, or if they can only capture some local object information. In this study, we find that the intermediate layers of models can encode more global semantic information, whose representation vectors perform better on visual-language entailment tasks, rather than the topmost layers. We further probe models for local semantic representation through object detection tasks. And we draw a conclusion that the topmost layers may excessively focus on local information, leading to a diminished ability to encode global information.
In light of recent advances in multimodal Large Language Models (LLMs), there is increasing attention to scaling them from image-text data to more informative real-world videos. Compared to static images, video poses unique challenges for effective large-scale pre-training due to the modeling of its spatiotemporal dynamics. In this paper, we address such limitations in video-language pre-training with an efficient video decomposition that represents each video as keyframes and temporal motions. These are then adapted to an LLM using well-designed tokenizers that discretize visual and temporal information as a few tokens, thus enabling unified generative pre-training of videos, images, and text. At inference, the generated tokens from the LLM are carefully recovered to the original continuous pixel space to create various video content. Our proposed framework is both capable of comprehending and generating image and video content, as demonstrated by its competitive performance across 13 multimodal benchmarks in image and video understanding and generation. Our code and models will be available at https://video-lavit.github.io.
While large language models exhibit remarkable performance in the Question Answering task, they are susceptible to hallucinations. Challenges arise when these models grapple with understanding multi-hop relations in complex questions or lack the necessary knowledge for a comprehensive response. To address this issue, we introduce the "Decompose-and-Query" framework (D&Q). This framework guides the model to think and utilize external knowledge similar to ReAct, while also restricting its thinking to reliable information, effectively mitigating the risk of hallucinations. Experiments confirm the effectiveness of D&Q: On our ChitChatQA dataset, D&Q does not lose to ChatGPT in 67% of cases; on the HotPotQA question-only setting, D&Q achieved an F1 score of 59.6%. Our code is available at https://github.com/alkaidpku/DQ-ToolQA.
Unlike alphabetic languages, Chinese spelling and pronunciation are different. Both characters and pinyin take an important role in Chinese language understanding. In Chinese NLP tasks, we almost adopt characters or words as model input, and few works study how to use pinyin. However, pinyin is essential in many scenarios, such as error correction and fault tolerance for ASR-introduced errors. Most of these errors are caused by the same or similar pronunciation words, and we refer to this type of error as SSP(the same or similar pronunciation) errors for short. In this work, we explore various ways of using pinyin in pretraining models and propose a new pretraining method called PmBERT. Our method uses characters and pinyin in parallel for pretraining. Through delicate pretraining tasks, the characters and pinyin representation are fused, which can enhance the error tolerance for SSP errors. We do comprehensive experiments and ablation tests to explore what makes a robust phonetic enhanced Chinese language model. The experimental results on both the constructed noise-added dataset and the public error-correction dataset demonstrate that our model is more robust compared to SOTA models.
Recently, the remarkable advance of the Large Language Model (LLM) has inspired researchers to transfer its extraordinary reasoning capability to both vision and language data. However, the prevailing approaches primarily regard the visual input as a prompt and focus exclusively on optimizing the text generation process conditioned upon vision content by a frozen LLM. Such an inequitable treatment of vision and language heavily constrains the model's potential. In this paper, we break through this limitation by representing both vision and language in a unified form. Specifically, we introduce a well-designed visual tokenizer to translate the non-linguistic image into a sequence of discrete tokens like a foreign language that LLM can read. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image. Coped with this tokenizer, the presented foundation model called LaVIT can handle both image and text indiscriminately under the same generative learning paradigm. This unification empowers LaVIT to serve as an impressive generalist interface to understand and generate multi-modal content simultaneously. Extensive experiments further showcase that it outperforms the existing models by a large margin on massive vision-language tasks. Our code and models will be available at https://github.com/jy0205/LaVIT.
Recently, the remarkable advance of the Large Language Model (LLM) has inspired researchers to transfer its extraordinary reasoning capability to data across several modalities. The prevailing approaches primarily regard visual input as the prompt and focus exclusively on optimizing the text generation process conditioned upon vision content by a frozen LLM. Such an inequitable treatment of vision and language heavily constrains the model's potential. In this paper, we break through this limitation by representing both vision and language in a unified representation. To this end, we craft a visual tokenizer that translates the non-linguistic image into a sequence of discrete tokens like a foreign language that LLM can read. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image content. Coped with this visual tokenizer, the presented foundation model called LaVIT (Language-VIsion Transformer) can handle both image and text indiscriminately under a unified generative learning paradigm. Pre-trained on the web-scale image-text corpus, LaVIT is empowered with impressive multi-modal comprehension capability. The extensive experiments showcase that it outperforms existing models by a large margin on downstream tasks. Our code and models will be available at https://github.com/jy0205/LaVIT.
The current optical communication systems minimize bit or symbol errors without considering the semantic meaning behind digital bits, thus transmitting a lot of unnecessary information. We propose and experimentally demonstrate a semantic optical fiber communication (SOFC) system. Instead of encoding information into bits for transmission, semantic information is extracted from the source using deep learning. The generated semantic symbols are then directly transmitted through an optical fiber. Compared with the bit-based structure, the SOFC system achieved higher information compression and a more stable performance, especially in the low received optical power regime, and enhanced the robustness against optical link impairments. This work introduces an intelligent optical communication system at the human analytical thinking level, which is a significant step toward a breakthrough in the current optical communication architecture.
Neural Radiance Field (NeRF) is a powerful tool to faithfully generate novel views for scenes with only sparse captured images. Despite its strong capability for representing 3D scenes and their appearance, its editing ability is very limited. In this paper, we propose a simple but effective extension of vanilla NeRF, named PaletteNeRF, to enable efficient color editing on NeRF-represented scenes. Motivated by recent palette-based image decomposition works, we approximate each pixel color as a sum of palette colors modulated by additive weights. Instead of predicting pixel colors as in vanilla NeRFs, our method predicts additive weights. The underlying NeRF backbone could also be replaced with more recent NeRF models such as KiloNeRF to achieve real-time editing. Experimental results demonstrate that our method achieves efficient, view-consistent, and artifact-free color editing on a wide range of NeRF-represented scenes.