Recent developments in large-scale pre-trained text-to-image diffusion models have significantly improved the generation of high-fidelity images, particularly with the emergence of diffusion models based on transformer architecture (DiTs). Among these diffusion models, diffusion transformers have demonstrated superior image generation capabilities, boosting lower FID scores and higher scalability. However, deploying large-scale DiT models can be expensive due to their extensive parameter numbers. Although existing research has explored efficient deployment techniques for diffusion models such as model quantization, there is still little work concerning DiT-based models. To tackle this research gap, in this paper, we propose TerDiT, a quantization-aware training (QAT) and efficient deployment scheme for ternary diffusion models with transformers. We focus on the ternarization of DiT networks and scale model sizes from 600M to 4.2B. Our work contributes to the exploration of efficient deployment strategies for large-scale DiT models, demonstrating the feasibility of training extremely low-bit diffusion transformer models from scratch while maintaining competitive image generation capacities compared to full-precision models. Code will be available at https://github.com/Lucky-Lance/TerDiT.
Text-Centric Visual Question Answering (TEC-VQA) in its proper format not only facilitates human-machine interaction in text-centric visual environments but also serves as a de facto gold proxy to evaluate AI models in the domain of text-centric scene understanding. However, most TEC-VQA benchmarks have focused on high-resource languages like English and Chinese. Despite pioneering works to expand multilingual QA pairs in non-text-centric VQA datasets using translation engines, the translation-based protocol encounters a substantial ``Visual-textual misalignment'' problem when applied to TEC-VQA. Specifically, it prioritizes the text in question-answer pairs while disregarding the visual text present in images. Furthermore, it does not adequately tackle challenges related to nuanced meaning, contextual distortion, language bias, and question-type diversity. In this work, we address the task of multilingual TEC-VQA and provide a benchmark with high-quality human expert annotations in 9 diverse languages, called MTVQA. To our knowledge, MTVQA is the first multilingual TEC-VQA benchmark to provide human expert annotations for text-centric scenarios. Further, by evaluating several state-of-the-art Multimodal Large Language Models (MLLMs), including GPT-4V, on our MTVQA dataset, it is evident that there is still room for performance improvement, underscoring the value of our dataset. We hope this dataset will provide researchers with fresh perspectives and inspiration within the community. The MTVQA dataset will be available at https://huggingface.co/datasets/ByteDance/MTVQA.
Click-Through Rate (CTR) prediction holds a paramount position in recommender systems. The prevailing ID-based paradigm underperforms in cold-start scenarios due to the skewed distribution of feature frequency. Additionally, the utilization of a single modality fails to exploit the knowledge contained within textual features. Recent efforts have sought to mitigate these challenges by integrating Pre-trained Language Models (PLMs). They design hard prompts to structure raw features into text for each interaction and then apply PLMs for text processing. With external knowledge and reasoning capabilities, PLMs extract valuable information even in cases of sparse interactions. Nevertheless, compared to ID-based models, pure text modeling degrades the efficacy of collaborative filtering, as well as feature scalability and efficiency during both training and inference. To address these issues, we propose \textbf{C}ost-\textbf{E}fficient \textbf{L}anguage Model \textbf{A}lignment (\textbf{CELA}) for CTR prediction. CELA incorporates textual features and language models while preserving the collaborative filtering capabilities of ID-based models. This model-agnostic framework can be equipped with plug-and-play textual features, with item-level alignment enhancing the utilization of external information while maintaining training and inference efficiency. Through extensive offline experiments, CELA demonstrates superior performance compared to state-of-the-art methods. Furthermore, an online A/B test conducted on an industrial App recommender system showcases its practical effectiveness, solidifying the potential for real-world applications of CELA.
Retrieval-Augmented Generation (RAG) has emerged as a pivotal innovation in natural language processing, enhancing generative models by incorporating external information retrieval. Evaluating RAG systems, however, poses distinct challenges due to their hybrid structure and reliance on dynamic knowledge sources. We consequently enhanced an extensive survey and proposed an analysis framework for benchmarks of RAG systems, RAGR (Retrieval, Generation, Additional Requirement), designed to systematically analyze RAG benchmarks by focusing on measurable outputs and established truths. Specifically, we scrutinize and contrast multiple quantifiable metrics of the Retrieval and Generation component, such as relevance, accuracy, and faithfulness, of the internal links within the current RAG evaluation methods, covering the possible output and ground truth pairs. We also analyze the integration of additional requirements of different works, discuss the limitations of current benchmarks, and propose potential directions for further research to address these shortcomings and advance the field of RAG evaluation. In conclusion, this paper collates the challenges associated with RAG evaluation. It presents a thorough analysis and examination of existing methodologies for RAG benchmark design based on the proposed RGAR framework.
Recent progress in large-scale scene rendering has yielded Neural Radiance Fields (NeRF)-based models with an impressive ability to synthesize scenes across small objects and indoor scenes. Nevertheless, extending this idea to large-scale aerial rendering poses two critical problems. Firstly, a single NeRF cannot render the entire scene with high-precision for complex large-scale aerial datasets since the sampling range along each view ray is insufficient to cover buildings adequately. Secondly, traditional NeRFs are infeasible to train on one GPU to enable interactive fly-throughs for modeling massive images. Instead, existing methods typically separate the whole scene into multiple regions and train a NeRF on each region, which are unaccustomed to different flight trajectories and difficult to achieve fast rendering. To that end, we propose Aerial-NeRF with three innovative modifications for jointly adapting NeRF in large-scale aerial rendering: (1) Designing an adaptive spatial partitioning and selection method based on drones' poses to adapt different flight trajectories; (2) Using similarity of poses instead of (expert) network for rendering speedup to determine which region a new viewpoint belongs to; (3) Developing an adaptive sampling approach for rendering performance improvement to cover the entire buildings at different heights. Extensive experiments have conducted to verify the effectiveness and efficiency of Aerial-NeRF, and new state-of-the-art results have been achieved on two public large-scale aerial datasets and presented SCUTic dataset. Note that our model allows us to perform rendering over 4 times as fast as compared to multiple competitors. Our dataset, code, and model are publicly available at https://drliuqi.github.io/.
Being able to successfully determine whether the testing samples has similar distribution as the training samples is a fundamental question to address before we can safely deploy most of the machine learning models into practice. In this paper, we propose TOOD detection, a simple yet effective tree-based out-of-distribution (TOOD) detection mechanism to determine if a set of unseen samples will have similar distribution as of the training samples. The TOOD detection mechanism is based on computing pairwise hamming distance of testing samples' tree embeddings, which are obtained by fitting a tree-based ensemble model through in-distribution training samples. Our approach is interpretable and robust for its tree-based nature. Furthermore, our approach is efficient, flexible to various machine learning tasks, and can be easily generalized to unsupervised setting. Extensive experiments are conducted to show the proposed method outperforms other state-of-the-art out-of-distribution detection methods in distinguishing the in-distribution from out-of-distribution on various tabular, image, and text data.
We introduce Vibe-Eval: a new open benchmark and framework for evaluating multimodal chat models. Vibe-Eval consists of 269 visual understanding prompts, including 100 of hard difficulty, complete with gold-standard responses authored by experts. Vibe-Eval is open-ended and challenging with dual objectives: (i) vibe checking multimodal chat models for day-to-day tasks and (ii) rigorously testing and probing the capabilities of present frontier models. Notably, our hard set contains >50% questions that all frontier models answer incorrectly. We explore the nuances of designing, evaluating, and ranking models on ultra challenging prompts. We also discuss trade-offs between human and automatic evaluation, and show that automatic model evaluation using Reka Core roughly correlates to human judgment. We offer free API access for the purpose of lightweight evaluation and plan to conduct formal human evaluations for public models that perform well on the Vibe-Eval's automatic scores. We release the evaluation code and data, see https://github.com/reka-ai/reka-vibe-eval
Text-centric visual question answering (VQA) has made great strides with the development of Multimodal Large Language Models (MLLMs), yet open-source models still fall short of leading models like GPT4V and Gemini, partly due to a lack of extensive, high-quality instruction tuning data. To this end, we introduce a new approach for creating a massive, high-quality instruction-tuning dataset, Square-10M, which is generated using closed-source MLLMs. The data construction process, termed Square, consists of four steps: Self-Questioning, Answering, Reasoning, and Evaluation. Our experiments with Square-10M led to three key findings: 1) Our model, TextSquare, considerably surpasses open-source previous state-of-the-art Text-centric MLLMs and sets a new standard on OCRBench(62.2%). It even outperforms top-tier models like GPT4V and Gemini in 6 of 10 text-centric benchmarks. 2) Additionally, we demonstrate the critical role of VQA reasoning data in offering comprehensive contextual insights for specific questions. This not only improves accuracy but also significantly mitigates hallucinations. Specifically, TextSquare scores an average of 75.1% across four general VQA and hallucination evaluation datasets, outperforming previous state-of-the-art models. 3) Notably, the phenomenon observed in scaling text-centric VQA datasets reveals a vivid pattern: the exponential increase of instruction tuning data volume is directly proportional to the improvement in model performance, thereby validating the necessity of the dataset scale and the high quality of Square-10M.
We introduce Reka Core, Flash, and Edge, a series of powerful multimodal language models trained from scratch by Reka. Reka models are able to process and reason with text, images, video, and audio inputs. This technical report discusses details of training some of these models and provides comprehensive evaluation results. We show that Reka Edge and Reka Flash are not only state-of-the-art but also outperform many much larger models, delivering outsized values for their respective compute class. Meanwhile, our most capable and largest model, Reka Core, approaches the best frontier models on both automatic evaluations and blind human evaluations. On image question answering benchmarks (e.g. MMMU, VQAv2), Core performs competitively to GPT4-V. Meanwhile, on multimodal chat, Core ranks as the second most preferred model under a blind third-party human evaluation setup, outperforming other models such as Claude 3 Opus. On text benchmarks, Core not only performs competitively to other frontier models on a set of well-established benchmarks (e.g. MMLU, GSM8K) but also outperforms GPT4-0613 on human evaluation. On video question answering (Perception-Test), Core outperforms Gemini Ultra. Models are shipped in production at http://chat.reka.ai . A showcase of non cherry picked qualitative examples can also be found at http://showcase.reka.ai .
Existing neural radiance fields (NeRF)-based novel view synthesis methods for large-scale outdoor scenes are mainly built on a single altitude. Moreover, they often require a priori camera shooting height and scene scope, leading to inefficient and impractical applications when camera altitude changes. In this work, we propose an end-to-end framework, termed AG-NeRF, and seek to reduce the training cost of building good reconstructions by synthesizing free-viewpoint images based on varying altitudes of scenes. Specifically, to tackle the detail variation problem from low altitude (drone-level) to high altitude (satellite-level), a source image selection method and an attention-based feature fusion approach are developed to extract and fuse the most relevant features of target view from multi-height images for high-fidelity rendering. Extensive experiments demonstrate that AG-NeRF achieves SOTA performance on 56 Leonard and Transamerica benchmarks and only requires a half hour of training time to reach the competitive PSNR as compared to the latest BungeeNeRF.