Author name cluster

Yufeng Cui

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

6 papers

2 author rows

TMLR Journal 2026 Journal Article

MIRA: Multi-view Information Retrieval with Adaptive Routing for Test-time Long-video Comprehension

Zecheng Hao
Wayne Ma
Yufeng Cui
Shuang Li
Xinlong Wang
Tiejun Huang

Foundational Multi-modal Large Language Models (MLLMs) have achieved rapid progress in handling complex tasks across diverse modalities. However, they still struggle to deliver satisfactory performance on Long-video Comprehension (LVC) tasks involving thousands of frames. Existing optimization strategies can be broadly categorized into LVC-specific fine-tuning, built-in token compression and training-free keyframe extraction, with the latter being most suitable for flexible deployment across various MLLMs. Unfortunately, current training-free approaches predominantly focus on query-frame relevance retrieval, overlooking other levels of visual information and the inherent heterogeneity of LVC tasks. In this work, we propose the $\textbf{M}$ulti-view $\textbf{I}$nformation $\textbf{R}$etrieval with $\textbf{A}$daptive Routing ($\textbf{MIRA}$) framework, which evaluates video frames using distinct metrics for relevance and causality, combines these scores to select a balanced pool of keyframes, and employs an adaptive feedback loop to tailor the retrieval process to different user queries, enabling more precise and sample-grained video comprehension. Extensive experiments demonstrate the advanced performance of our scheme across multiple challenging LVC benchmarks. For instance, integrating $\textbf{MIRA}$ with Qwen-2.5-VL yields performance gains of 3.5% to 13.1% on LVB, VideoMME and MLVU.

PDF Details

ICLR Conference 2025 Conference Paper

Autoregressive Video Generation without Vector Quantization

Haoge Deng
Ting Pan
Haiwen Diao
Zhengxiong Luo
Yufeng Cui
Huchuan Lu
Shiguang Shan
Yonggang Qi

This paper presents a novel approach that enables autoregressive video generation with high efficiency. We propose to reformulate the video generation problem as a non-quantized autoregressive modeling of temporal frame-by-frame prediction and spatial set-by-set prediction. Unlike raster-scan prediction in prior autoregressive models or joint distribution modeling of fixed-length tokens in diffusion models, our approach maintains the causal property of GPT-style models for flexible in-context capabilities, while leveraging bidirectional modeling within individual frames for efficiency. With the proposed approach, we train a novel video autoregressive model without vector quantization, termed NOVA. Our results demonstrate that NOVA surpasses prior autoregressive video models in data efficiency, inference speed, visual fidelity, and video fluency, even with a much smaller model capacity, i.e., 0.6B parameters. NOVA also outperforms state-of-the-art image diffusion models in text-to-image generation tasks, with a significantly lower training cost. Additionally, NOVA generalizes well across extended video durations and enables diverse zero-shot applications in one unified model. Code and models are publicly available at https://github.com/baaivision/NOVA.

Details

NeurIPS Conference 2025 Conference Paper

End-to-End Vision Tokenizer Tuning

Wenxuan Wang
Fan Zhang
Yufeng Cui
Haiwen Diao
Zhuoyan Luo
Huchuan Lu
Jing Liu
Xinlong Wang

Existing vision tokenization isolates the optimization of vision tokenizers from downstream training, implicitly assuming the visual tokens can generalize well across various tasks, e. g. , image generation and visual question answering. The vision tokenizer optimized for low-level reconstruction is agnostic to downstream tasks requiring varied representations and semantics. This decoupled paradigm introduces a critical misalignment: The loss of the vision tokenization can be the representation bottleneck for target tasks. For example, errors in tokenizing text in a given image lead to poor results when recognizing or generating them. To address this, we propose ETT, an end-to-end vision tokenizer tuning approach that enables joint optimization between vision tokenization and target autoregressive tasks. Unlike prior autoregressive models that use only discrete indices from a frozen vision tokenizer, ETT leverages the visual embeddings of the tokenizer codebook, and optimizes the vision tokenizers end-to-end with both reconstruction and caption objectives. ETT can be seamlessly integrated into existing training pipelines with minimal architecture modifications. Our ETT is simple to implement and integrate, without the need to adjust the original codebooks or architectures of the employed large language models. Extensive experiments demonstrate that our proposed end-to-end vision tokenizer tuning unlocks significant performance gains, i. e. , 2-6% for multimodal understanding and visual generation tasks compared to frozen tokenizer baselines, while preserving the original reconstruction capability. We hope this very simple and strong method can empower multimodal foundation models besides image generation and understanding.

PDF Details

ICLR Conference 2024 Conference Paper

Emu: Generative Pretraining in Multimodality

Quan Sun
Qiying Yu
Yufeng Cui
Fan Zhang
Xiaosong Zhang
Yueze Wang
Hongcheng Gao
Jingjing Liu

We present Emu, a multimodal foundation model that seamlessly generates images and text in multimodal context. This omnivore model can take in any single-modality or multimodal data input indiscriminately (e.g., interleaved image, text and video) through a one-model-for-all autoregressive training process. First, visual signals are encoded into embeddings, and together with text tokens form an interleaved input sequence. Emu is end-to-end trained with a unified objective of classifying the next text token or regressing the next visual embedding in the multimodal sequence. This versatile multimodality empowers the leverage of diverse pretraining data sources at scale, such as videos with interleaved frames and text, webpages with interleaved images and text, as well as web-scale image-text pairs and video-text pairs. Emu can serve as a generalist multimodal interface for both image-to-text and text-to-image tasks, supporting in-context image and text generation. Across a broad range of zero-shot/few-shot tasks including image captioning, visual question answering, video question answering and text-to-image generation, Emu demonstrates superb performance compared to state-of-the-art large multimodal models. Extended capabilities such as multimodal assistants via instruction tuning are also demonstrated with impressive performance.

Details

NeurIPS Conference 2024 Conference Paper

Unveiling Encoder-Free Vision-Language Models

Haiwen Diao
Yufeng Cui
Xiaotong Li
Yueze Wang
Huchuan Lu
Xinlong Wang

Existing vision-language models (VLMs) mostly rely on vision encoders to extract visual features followed by large language models (LLMs) for visual-language tasks. However, the vision encoders set a strong inductive bias in abstracting visual representation, e. g. , resolution, aspect ratio, and semantic priors, which could impede the flexibility and efficiency of the VLMs. Training pure VLMs that accept the seamless vision and language inputs, i. e. , without vision encoders, remains challenging and rarely explored. Empirical observations reveal that direct training without encoders results in slow convergence and large performance gaps. In this work, we bridge the gap between encoder-based and encoder-free models, and present a simple yet effective training recipe towards pure VLMs. Specifically, we unveil the key aspects of training encoder-free VLMs efficiently via thorough experiments: (1) Bridging vision-language representation inside one unified decoder; (2) Enhancing visual recognition capability via extra supervision. With these strategies, we launch EVE, an encoder-free vision-language model that can be trained and forwarded efficiently. Notably, solely utilizing 35M publicly accessible data, EVE can impressively rival the encoder-based VLMs of similar capacities across multiple vision-language benchmarks. It significantly outperforms the counterpart Fuyu-8B with mysterious training procedures and undisclosed training data. We believe that EVE provides a transparent and efficient route for developing pure decoder-only architecture across modalities.

PDF Details DOI

ICLR Conference 2022 Conference Paper

Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm

Yangguang Li 0001
Feng Liang
Lichen Zhao
Yufeng Cui
Wanli Ouyang
Jing Shao
Fengwei Yu
Junjie Yan

Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks. However, CLIP is quite data-hungry and requires 400M image-text pairs for pre-training, thereby restricting its adoption. This work proposes a novel training paradigm, Data efficient CLIP (DeCLIP), to alleviate this limitation. We demonstrate that by carefully utilizing the widespread supervision among the image-text pairs, our De-CLIP can learn generic visual features more efficiently. Instead of using the single image-text contrastive supervision, we fully exploit data potential through the use of (1) self-supervision within each modality; (2) multi-view supervision across modalities; (3) nearest-neighbor supervision from other similar pairs. Benefiting from intrinsic supervision, our DeCLIP-ResNet50 can achieve 60.4% zero-shot top1 accuracy on ImageNet, which is 0.8% above the CLIP-ResNet50 while using 7.1×fewer data. Our DeCLIP-ResNet50 outperforms its counterpart in 8 out of 11 visual datasets when transferred to downstream tasks. Moreover, Scaling up the model and computing also works well in our framework.

Details