Junnan Li

Anthony Meng Huat Tiong, Junnan Li, Boyang Li et al.

Visual question answering (VQA) is a hallmark of vision and language reasoning and a challenging task under the zero-shot setting. We propose Plug-and-Play VQA (PNP-VQA), a modular framework for zero-shot VQA. In contrast to most existing works, which require substantial adaptation of pretrained language models (PLMs) for the vision modality, PNP-VQA requires no additional training of the PLMs. Instead, we propose to use natural language and network interpretation as an intermediate representation that glues pretrained models together. We first generate question-guided informative image captions, and pass the captions to a PLM as context for question answering. Surpassing end-to-end trained baselines, PNP-VQA achieves state-of-the-art results on zero-shot VQAv2 and GQA. With 11B parameters, it outperforms the 80B-parameter Flamingo model by 8.5% on VQAv2. With 738M PLM parameters, PNP-VQA achieves an improvement of 9.1% on GQA over FewVLM with 740M PLM parameters. Code is released at https://github.com/salesforce/LAVIS/tree/main/projects/pnp-vqa

Dongxu Li, Junnan Li, Hung Le et al.

We introduce LAVIS, an open-source deep learning library for LAnguage-VISion research and applications. LAVIS aims to serve as a one-stop comprehensive library that brings recent advancements in the language-vision field accessible for researchers and practitioners, as well as fertilizing future research and development. It features a unified interface to easily access state-of-the-art image-language, video-language models and common datasets. LAVIS supports training, evaluation and benchmarking on a rich variety of tasks, including multimodal classification, retrieval, captioning, visual question answering, dialogue and pre-training. In the meantime, the library is also highly extensible and configurable, facilitating future development and customization. In this technical report, we describe design principles, key components and functionalities of the library, and also present benchmarking results across common language-vision tasks. The library is available at: https://github.com/salesforce/LAVIS.

Artemis Panagopoulou, Le Xue, Ning Yu et al. · salesforce, stanford

Recent research has achieved significant advancements in visual reasoning tasks through learning image-to-language projections and leveraging the impressive reasoning abilities of Large Language Models (LLMs). This paper introduces an efficient and effective framework that integrates multiple modalities (images, 3D, audio and video) to a frozen LLM and demonstrates an emergent ability for cross-modal reasoning (2+ modality inputs). Our approach explores two distinct projection mechanisms: Q-Formers and Linear Projections (LPs). Through extensive experimentation across all four modalities on 16 benchmarks, we explore both methods and assess their adaptability in integrated and separate cross-modal reasoning. The Q-Former projection demonstrates superior performance in single modality scenarios and adaptability in joint versus discriminative reasoning involving two or more modalities. However, it exhibits lower generalization capabilities than linear projection in contexts where task-modality data are limited. To enable this framework, we devise a scalable pipeline that automatically generates high-quality, instruction-tuning datasets from readily available captioning data across different modalities, and contribute 24K QA data for audio and 250K QA data for 3D. To facilitate further research in cross-modal reasoning, we introduce the DisCRn (Discriminative Cross-modal Reasoning) benchmark comprising 9K audio-video QA samples and 28K image-3D QA samples that require the model to reason discriminatively across disparate input modalities.

Haoning Wu, Dongxu Li, Bei Chen et al.

Large multimodal models (LMMs) are processing increasingly longer and richer inputs. Albeit the progress, few public benchmark is available to measure such development. To mitigate this gap, we introduce LongVideoBench, a question-answering benchmark that features video-language interleaved inputs up to an hour long. Our benchmark includes 3,763 varying-length web-collected videos with their subtitles across diverse themes, designed to comprehensively evaluate LMMs on long-term multimodal understanding. To achieve this, we interpret the primary challenge as to accurately retrieve and reason over detailed multimodal information from long inputs. As such, we formulate a novel video question-answering task termed referring reasoning. Specifically, as part of the question, it contains a referring query that references related video contexts, called referred context. The model is then required to reason over relevant video details from the referred context. Following the paradigm of referring reasoning, we curate 6,678 human-annotated multiple-choice questions in 17 fine-grained categories, establishing one of the most comprehensive benchmarks for long-form video understanding. Evaluations suggest that the LongVideoBench presents significant challenges even for the most advanced proprietary models (e.g. GPT-4o, Gemini-1.5-Pro, GPT-4-Turbo), while their open-source counterparts show an even larger performance gap. In addition, our results indicate that model performance on the benchmark improves only when they are capable of processing more frames, positioning LongVideoBench as a valuable benchmark for evaluating future-generation long-context LMMs.

Junnan Li, Dongxu Li, Silvio Savarese et al.

The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions.

Guangsen Wang, Shafiq Joty, Junnan Li et al.

We introduce BotSIM, a modular, open-source Bot SIMulation environment with dialog generation, user simulation and conversation analytics capabilities. BotSIM aims to serve as a one-stop solution for large-scale data-efficient end-to-end evaluation, diagnosis and remediation of commercial task-oriented dialog (TOD) systems to significantly accelerate commercial bot development and evaluation, reduce cost and time-to-market. BotSIM adopts a layered design comprising the infrastructure layer, the adaptor layer and the application layer. The infrastructure layer hosts key models and components to support BotSIM's major functionalities via a streamlined "generation-simulation-remediation" pipeline. The adaptor layer is used to extend BotSIM to accommodate new bot platforms. The application layer provides a suite of command line tools and a Web App to significantly lower the entry barrier for BotSIM users such as bot admins or practitioners. In this report, we focus on the technical designs of various system components. A detailed case study using Einstein BotBuilder is also presented to show how to apply BotSIM pipeline for bot evaluation and remediation. The detailed system descriptions can be found in our system demo paper. The toolkit is available at: https://github.com/salesforce/BotSIM .

Jiaxian Guo, Junnan Li, Dongxu Li et al.

Large language models (LLMs) have demonstrated excellent zero-shot generalization to new language tasks. However, effective utilization of LLMs for zero-shot visual question-answering (VQA) remains challenging, primarily due to the modality disconnection and task disconnection between LLM and VQA task. End-to-end training on vision and language data may bridge the disconnections, but is inflexible and computationally expensive. To address this issue, we propose \emph{Img2Prompt}, a plug-and-play module that provides the prompts that can bridge the aforementioned modality and task disconnections, so that LLMs can perform zero-shot VQA tasks without end-to-end training. In order to provide such prompts, we further employ LLM-agnostic models to provide prompts that can describe image content and self-constructed question-answer pairs, which can effectively guide LLM to perform zero-shot VQA tasks. Img2Prompt offers the following benefits: 1) It can flexibly work with various LLMs to perform VQA. 2)~Without the needing of end-to-end training, it significantly reduces the cost of deploying LLM for zero-shot VQA tasks. 3) It achieves comparable or better performance than methods relying on end-to-end training. For example, we outperform Flamingo \cite{Deepmind:Flamingo2022} by 5.6\% on VQAv2. On the challenging A-OKVQA dataset, our method even outperforms few-shot methods by as much as 20\%.

Zhiyuan Hu, Yibo Wang, Hanze Dong et al. · salesforce

Large reasoning models (LRMs) already possess a latent capacity for long chain-of-thought reasoning. Prior work has shown that outcome-based reinforcement learning (RL) can incidentally elicit advanced reasoning behaviors such as self-correction, backtracking, and verification phenomena often referred to as the model's "aha moment". However, the timing and consistency of these emergent behaviors remain unpredictable and uncontrollable, limiting the scalability and reliability of LRMs' reasoning capabilities. To address these limitations, we move beyond reliance on prompts and coincidental "aha moments". Instead, we explicitly align models with three meta-abilities: deduction, induction, and abduction, using automatically generated, self-verifiable tasks. Our three stage-pipeline individual alignment, parameter-space merging, and domain-specific reinforcement learning, boosting performance by over 10\% relative to instruction-tuned baselines. Furthermore, domain-specific RL from the aligned checkpoint yields an additional gain in performance ceiling for both 7B and 32B models across math, coding, and science benchmarks, demonstrating that explicit meta-ability alignment offers a scalable and dependable foundation for reasoning. Code is available at: https://github.com/zhiyuanhubj/Meta-Ability-Alignment

Jiuhai Chen, Le Xue, Zhiyang Xu et al.

We present BLIP3o-NEXT, a fully open-source foundation model in the BLIP3 series that advances the next frontier of native image generation. BLIP3o-NEXT unifies text-to-image generation and image editing within a single architecture, demonstrating strong image generation and image editing capabilities. In developing the state-of-the-art native image generation model, we identify four key insights: (1) Most architectural choices yield comparable performance; an architecture can be deemed effective provided it scales efficiently and supports fast inference; (2) The successful application of reinforcement learning can further push the frontier of native image generation; (3) Image editing still remains a challenging task, yet instruction following and the consistency between generated and reference images can be significantly enhanced through post-training and data engine; (4) Data quality and scale continue to be decisive factors that determine the upper bound of model performance. Building upon these insights, BLIP3o-NEXT leverages an Autoregressive + Diffusion architecture in which an autoregressive model first generates discrete image tokens conditioned on multimodal inputs, whose hidden states are then used as conditioning signals for a diffusion model to generate high-fidelity images. This architecture integrates the reasoning strength and instruction following of autoregressive models with the fine-detail rendering ability of diffusion models, achieving a new level of coherence and realism. Extensive evaluations of various text-to-image and image-editing benchmarks show that BLIP3o-NEXT achieves superior performance over existing models.

Dongxu Li, Junnan Li, Steven C. H. Hoi

Subject-driven text-to-image generation models create novel renditions of an input subject based on text prompts. Existing models suffer from lengthy fine-tuning and difficulties preserving the subject fidelity. To overcome these limitations, we introduce BLIP-Diffusion, a new subject-driven image generation model that supports multimodal control which consumes inputs of subject images and text prompts. Unlike other subject-driven generation models, BLIP-Diffusion introduces a new multimodal encoder which is pre-trained to provide subject representation. We first pre-train the multimodal encoder following BLIP-2 to produce visual representation aligned with the text. Then we design a subject representation learning task which enables a diffusion model to leverage such visual representation and generates new subject renditions. Compared with previous methods such as DreamBooth, our model enables zero-shot subject-driven generation, and efficient fine-tuning for customized subject with up to 20x speedup. We also demonstrate that BLIP-Diffusion can be flexibly combined with existing techniques such as ControlNet and prompt-to-prompt to enable novel subject-driven generation and editing applications. Code and models will be released at https://github.com/salesforce/LAVIS/tree/main/projects/blip-diffusion. Project page at https://dxli94.github.io/BLIP-Diffusion-website/.

Wenliang Dai, Junnan Li, Dongxu Li et al.

Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip.

Junnan Li, Dongxu Li, Caiming Xiong et al.

Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP.

Dongxu Li, Junnan Li, Hongdong Li et al.

Video-and-language pre-training has shown promising improvements on various downstream tasks. Most previous methods capture cross-modal interactions with a transformer-based multimodal encoder, not fully addressing the misalignment between unimodal video and text features. Besides, learning fine-grained visual-language alignment usually requires off-the-shelf object detectors to provide object information, which is bottlenecked by the detector's limited vocabulary and expensive computation cost. We propose Align and Prompt: an efficient and effective video-and-language pre-training framework with better cross-modal alignment. First, we introduce a video-text contrastive (VTC) loss to align unimodal video-text features at the instance level, which eases the modeling of cross-modal interactions. Then, we propose a new visually-grounded pre-training task, prompting entity modeling (PEM), which aims to learn fine-grained region-entity alignment. To achieve this, we first introduce an entity prompter module, which is trained with VTC to produce the similarity between a video crop and text prompts instantiated with entity names. The PEM task then asks the model to predict the entity pseudo-labels (i.e~normalized similarity scores) for randomly-selected video crops. The resulting pre-trained model achieves state-of-the-art performance on both text-video retrieval and videoQA, outperforming prior work by a substantial margin. Our code and pre-trained models are available at https://github.com/salesforce/ALPRO.

Mingfei Gao, Chen Xing, Juan Carlos Niebles et al.

Despite great progress in object detection, most existing methods work only on a limited set of object categories, due to the tremendous human effort needed for bounding-box annotations of training data. To alleviate the problem, recent open vocabulary and zero-shot detection methods attempt to detect novel object categories beyond those seen during training. They achieve this goal by training on a pre-defined base categories to induce generalization to novel objects. However, their potential is still constrained by the small set of base categories available for training. To enlarge the set of base classes, we propose a method to automatically generate pseudo bounding-box annotations of diverse objects from large-scale image-caption pairs. Our method leverages the localization ability of pre-trained vision-language models to generate pseudo bounding-box labels and then directly uses them for training object detectors. Experimental results show that our method outperforms the state-of-the-art open vocabulary detector by 8% AP on COCO novel categories, by 6.3% AP on PASCAL VOC, by 2.3% AP on Objects365 and by 2.8% AP on LVIS. Code is available at https://github.com/salesforce/PB-OVD.

Junnan Li, Ramprasaath R. Selvaraju, Akhilesh Deepak Gotmare et al.

Large-scale vision and language representation learning has shown promising improvements on various vision-language tasks. Most existing methods employ a transformer-based multimodal encoder to jointly model visual tokens (region-based image features) and word tokens. Because the visual tokens and word tokens are unaligned, it is challenging for the multimodal encoder to learn image-text interactions. In this paper, we introduce a contrastive loss to ALign the image and text representations BEfore Fusing (ALBEF) them through cross-modal attention, which enables more grounded vision and language representation learning. Unlike most existing methods, our method does not require bounding box annotations nor high-resolution images. In order to improve learning from noisy web data, we propose momentum distillation, a self-training method which learns from pseudo-targets produced by a momentum model. We provide a theoretical analysis of ALBEF from a mutual information maximization perspective, showing that different training tasks can be interpreted as different ways to generate views for an image-text pair. ALBEF achieves state-of-the-art performance on multiple downstream vision-language tasks. On image-text retrieval, ALBEF outperforms methods that are pre-trained on orders of magnitude larger datasets. On VQA and NLVR$^2$, ALBEF achieves absolute improvements of 2.37% and 3.84% compared to the state-of-the-art, while enjoying faster inference speed. Code and pre-trained models are available at https://github.com/salesforce/ALBEF/.

Viraj Prabhu, Yutong Dai, Matthew Fernandez et al.

Web agents promise to automate complex browser tasks, but current methods remain brittle -- relying on step-by-step UI interactions and heavy LLM reasoning that break under dynamic layouts and long horizons. Humans, by contrast, exploit website-provided functionality through high-level operations like search, filter, and sort. We introduce WALT (Web Agents that Learn Tools), a framework that reverse-engineers latent website functionality into reusable invocable tools. Rather than hypothesizing ad-hoc skills, WALT exposes robust implementations of automations already designed into websites -- spanning discovery (search, filter, sort), communication (post, comment, upvote), and content management (create, edit, delete). Tools abstract away low-level execution: instead of reasoning about how to click and type, agents simply call search(query) or create(listing). This shifts the computational burden from fragile step-by-step reasoning to reliable tool invocation. On VisualWebArena and WebArena, WALT achieves higher success with fewer steps and less LLM-dependent reasoning, establishing a robust and generalizable paradigm for browser automation.

Akhilesh Deepak Gotmare, Junnan Li, Shafiq Joty et al.

The goal of natural language semantic code search is to retrieve a semantically relevant code snippet from a fixed set of candidates using a natural language query. Existing approaches are neither effective nor efficient enough towards a practical semantic code search system. In this paper, we propose an efficient and accurate semantic code search framework with cascaded fast and slow models, in which a fast transformer encoder model is learned to optimize a scalable index for fast retrieval followed by learning a slow classification-based re-ranking model to improve the performance of the top K results from the fast retrieval. To further reduce the high memory cost of deploying two separate models in practice, we propose to jointly train the fast and slow model based on a single transformer encoder with shared parameters. The proposed cascaded approach is not only efficient and scalable, but also achieves state-of-the-art results with an average mean reciprocal ranking (MRR) score of 0.7795 (across 6 programming languages) as opposed to the previous state-of-the-art result of 0.713 MRR on the CodeSearchNet benchmark.

Junnan Li, Yongkang Wong, Qi Zhao et al.

Social relationships form the basis of social structure of humans. Developing computational models to understand social relationships from visual data is essential for building intelligent machines that can better interact with humans in a social environment. In this work, we study the problem of visual social relationship recognition in images. We propose a Dual-Glance model for social relationship recognition, where the first glance fixates at the person of interest and the second glance deploys attention mechanism to exploit contextual cues. To enable this study, we curated a large scale People in Social Context (PISC) dataset, which comprises of 23,311 images and 79,244 person pairs with annotated social relationships. Since visually identifying social relationship bears certain degree of uncertainty, we further propose an Adaptive Focal Loss to leverage the ambiguous annotations for more effective learning. We conduct extensive experiments to quantitatively and qualitatively demonstrate the efficacy of our proposed method, which yields state-of-the-art performance on social relationship recognition.

Bingjie Xu, Junnan Li, Yongkang Wong et al.

The recent advances in instance-level detection tasks lay strong foundation for genuine comprehension of the visual scenes. However, the ability to fully comprehend a social scene is still in its preliminary stage. In this work, we focus on detecting human-object interactions (HOIs) in social scene images, which is demanding in terms of research and increasingly useful for practical applications. To undertake social tasks interacting with objects, humans direct their attention and move their body based on their intention. Based on this observation, we provide a unique computational perspective to explore human intention in HOI detection. Specifically, the proposed human intention-driven HOI detection (iHOI) framework models human pose with the relative distances from body joints to the object instances. It also utilizes human gaze to guide the attended contextual regions in a weakly-supervised setting. In addition, we propose a hard negative sampling strategy to address the problem of mis-grouping. We perform extensive experiments on two benchmark datasets, namely V-COCO and HICO-DET. The efficacy of each proposed component has also been validated.

Junnan Li, Yongkang Wong, Qi Zhao et al.

Bridging vision and natural language is a longstanding goal in computer vision and multimedia research. While earlier works focus on generating a single-sentence description for visual content, recent works have studied paragraph generation. In this work, we introduce the problem of video storytelling, which aims at generating coherent and succinct stories for long videos. Video storytelling introduces new challenges, mainly due to the diversity of the story and the length and complexity of the video. We propose novel methods to address the challenges. First, we propose a context-aware framework for multimodal embedding learning, where we design a Residual Bidirectional Recurrent Neural Network to leverage contextual information from past and future. Second, we propose a Narrator model to discover the underlying storyline. The Narrator is formulated as a reinforcement learning agent which is trained by directly optimizing the textual metric of the generated story. We evaluate our method on the Video Story dataset, a new dataset that we have collected to enable the study. We compare our method with multiple state-of-the-art baselines, and show that our method achieves better performance, in terms of quantitative measures and user study.

Junnan Li, Yongkang Wong, Qi Zhao et al.

Since the beginning of early civilizations, social relationships derived from each individual fundamentally form the basis of social structure in our daily life. In the computer vision literature, much progress has been made in scene understanding, such as object detection and scene parsing. Recent research focuses on the relationship between objects based on its functionality and geometrical relations. In this work, we aim to study the problem of social relationship recognition, in still images. We have proposed a dual-glance model for social relationship recognition, where the first glance fixates at the individual pair of interest and the second glance deploys attention mechanism to explore contextual cues. We have also collected a new large scale People in Social Context (PISC) dataset, which comprises of 22,670 images and 76,568 annotated samples from 9 types of social relationship. We provide benchmark results on the PISC dataset, and qualitatively demonstrate the efficacy of the proposed model.