MM-Embed: Universal Multimodal Retrieval with Multimodal LLMs

State-of-the-art retrieval models typically address a straightforward search scenario, where retrieval tasks are fixed (e.g., finding a passage to answer a specific question) and only a single modality is supported for both queries and retrieved results. This paper introduces techniques for advancing information retrieval with multimodal large language models (MLLMs), enabling a broader search scenario, termed universal multimodal retrieval, where multiple modalities and diverse retrieval tasks are accommodated. To this end, we first study fine-tuning an MLLM as a bi-encoder retriever on 10 datasets with 16 retrieval tasks. Our empirical results show that the fine-tuned MLLM retriever is capable of understanding challenging queries, composed of both text and image, but underperforms a smaller CLIP retriever in cross-modal retrieval tasks due to modality bias from MLLMs. To address the issue, we propose modality-aware hard negative mining to mitigate the modality bias exhibited by MLLM retrievers. Second, we propose to continually fine-tune the universal multimodal retriever to enhance its text retrieval capability while maintaining multimodal retrieval capability. As a result, our model, MM-Embed, achieves state-of-the-art performance on the multimodal retrieval benchmark M-BEIR, which spans multiple domains and tasks, while also surpassing the state-of-the-art text retrieval model, NV-Embed-v1, on MTEB retrieval benchmark. Finally, we explore to prompt the off-the-shelf MLLMs as the zero-shot rerankers to refine the ranking of the candidates from the multimodal retriever. We find that through prompt-and-reranking, MLLMs can further improve multimodal retrieval when the user queries (e.g., text-image composed queries) are more complex and challenging to understand. These findings also pave the way to advance universal multimodal retrieval in the future.

Synergistic Approach for Simultaneous Optimization of Monolingual, Cross-lingual, and Multilingual Information Retrieval

Information retrieval across different languages is an increasingly important challenge in natural language processing. Recent approaches based on multilingual pre-trained language models have achieved remarkable success, yet they often optimize for either monolingual, cross-lingual, or multilingual retrieval performance at the expense of others. This paper proposes a novel hybrid batch training strategy to simultaneously improve zero-shot retrieval performance across monolingual, cross-lingual, and multilingual settings while mitigating language bias. The approach fine-tunes multilingual language models using a mix of monolingual and cross-lingual question-answer pair batches sampled based on dataset size. Experiments on XQuAD-R, MLQA-R, and MIRACL benchmark datasets show that the proposed method consistently achieves comparable or superior results in zero-shot retrieval across various languages and retrieval tasks compared to monolingual-only or cross-lingual-only training. Hybrid batch training also substantially reduces language bias in multilingual retrieval compared to monolingual training. These results demonstrate the effectiveness of the proposed approach for learning language-agnostic representations that enable strong zero-shot retrieval performance across diverse languages.

Unifying Multimodal Retrieval via Document Screenshot Embedding

In the real world, documents are organized in different formats and varied modalities. Traditional retrieval pipelines require tailored document parsing techniques and content extraction modules to prepare input for indexing. This process is tedious, prone to errors, and has information loss. To this end, we propose Document Screenshot Embedding} (DSE), a novel retrieval paradigm that regards document screenshots as a unified input format, which does not require any content extraction preprocess and preserves all the information in a document (e.g., text, image and layout). DSE leverages a large vision-language model to directly encode document screenshots into dense representations for retrieval. To evaluate our method, we first craft the dataset of Wiki-SS, a 1.3M Wikipedia web page screenshots as the corpus to answer the questions from the Natural Questions dataset. In such a text-intensive document retrieval setting, DSE shows competitive effectiveness compared to other text retrieval methods relying on parsing. For example, DSE outperforms BM25 by 17 points in top-1 retrieval accuracy. Additionally, in a mixed-modality task of slide retrieval, DSE significantly outperforms OCR text retrieval methods by over 15 points in nDCG@10. These experiments show that DSE is an effective document retrieval paradigm for diverse types of documents. Model checkpoints, code, and Wiki-SS collection will be released.

FLAME: Factuality-Aware Alignment for Large Language Models

Alignment is a standard procedure to fine-tune pre-trained large language models (LLMs) to follow natural language instructions and serve as helpful AI assistants. We have observed, however, that the conventional alignment process fails to enhance the factual accuracy of LLMs, and often leads to the generation of more false facts (i.e. hallucination). In this paper, we study how to make the LLM alignment process more factual, by first identifying factors that lead to hallucination in both alignment steps:\ supervised fine-tuning (SFT) and reinforcement learning (RL). In particular, we find that training the LLM on new knowledge or unfamiliar texts can encourage hallucination. This makes SFT less factual as it trains on human labeled data that may be novel to the LLM. Furthermore, reward functions used in standard RL can also encourage hallucination, because it guides the LLM to provide more helpful responses on a diverse set of instructions, often preferring longer and more detailed responses. Based on these observations, we propose factuality-aware alignment, comprised of factuality-aware SFT and factuality-aware RL through direct preference optimization. Experiments show that our proposed factuality-aware alignment guides LLMs to output more factual responses while maintaining instruction-following capability.

Improving Conversational Passage Re-ranking with View Ensemble

This paper presents ConvRerank, a conversational passage re-ranker that employs a newly developed pseudo-labeling approach. Our proposed view-ensemble method enhances the quality of pseudo-labeled data, thus improving the fine-tuning of ConvRerank. Our experimental evaluation on benchmark datasets shows that combining ConvRerank with a conversational dense retriever in a cascaded manner achieves a good balance between effectiveness and efficiency. Compared to baseline methods, our cascaded pipeline demonstrates lower latency and higher top-ranking effectiveness. Furthermore, the in-depth analysis confirms the potential of our approach to improving the effectiveness of conversational search.

How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval

Various techniques have been developed in recent years to improve dense retrieval (DR), such as unsupervised contrastive learning and pseudo-query generation. Existing DRs, however, often suffer from effectiveness tradeoffs between supervised and zero-shot retrieval, which some argue was due to the limited model capacity. We contradict this hypothesis and show that a generalizable DR can be trained to achieve high accuracy in both supervised and zero-shot retrieval without increasing model size. In particular, we systematically examine the contrastive learning of DRs, under the framework of Data Augmentation (DA). Our study shows that common DA practices such as query augmentation with generative models and pseudo-relevance label creation using a cross-encoder, are often inefficient and sub-optimal. We hence propose a new DA approach with diverse queries and sources of supervision to progressively train a generalizable DR. As a result, DRAGON, our dense retriever trained with diverse augmentation, is the first BERT-base-sized DR to achieve state-of-the-art effectiveness in both supervised and zero-shot evaluations and even competes with models using more complex late interaction (ColBERTv2 and SPLADE++).

SLIM: Sparsified Late Interaction for Multi-Vector Retrieval with Inverted Indexes

This paper introduces a method called Sparsified Late Interaction for Multi-vector retrieval with inverted indexes (SLIM). Although multi-vector models have demonstrated their effectiveness in various information retrieval tasks, most of their pipelines require custom optimization to be efficient in both time and space. Among them, ColBERT is probably the most established method which is based on the late interaction of contextualized token embeddings of pre-trained language models. Unlike ColBERT where all its token embeddings are low-dimensional and dense, SLIM projects each token embedding into a high-dimensional, sparse lexical space before performing late interaction. In practice, we further propose to approximate SLIM using the lower- and upper-bound of the late interaction to reduce latency and storage. In this way, the sparse outputs can be easily incorporated into an inverted search index and are fully compatible with off-the-shelf search tools such as Pyserini and Elasticsearch. SLIM has competitive accuracy on information retrieval benchmarks such as MS MARCO Passages and BEIR compared to ColBERT while being much smaller and faster on CPUs. Source code and data will be available at https://github.com/castorini/pyserini/blob/master/docs/experiments-slim.md.

CITADEL: Conditional Token Interaction via Dynamic Lexical Routing for Efficient and Effective Multi-Vector Retrieval

Multi-vector retrieval methods combine the merits of sparse (e.g. BM25) and dense (e.g. DPR) retrievers and have achieved state-of-the-art performance on various retrieval tasks. These methods, however, are orders of magnitude slower and need much more space to store their indices compared to their single-vector counterparts. In this paper, we unify different multi-vector retrieval models from a token routing viewpoint and propose conditional token interaction via dynamic lexical routing, namely CITADEL, for efficient and effective multi-vector retrieval. CITADEL learns to route different token vectors to the predicted lexical ``keys'' such that a query token vector only interacts with document token vectors routed to the same key. This design significantly reduces the computation cost while maintaining high accuracy. Notably, CITADEL achieves the same or slightly better performance than the previous state of the art, ColBERT-v2, on both in-domain (MS MARCO) and out-of-domain (BEIR) evaluations, while being nearly 40 times faster. Code and data are available at https://github.com/facebookresearch/dpr-scale.

Strong Gravitational Lensing Parameter Estimation with Vision Transformer

Quantifying the parameters and corresponding uncertainties of hundreds of strongly lensed quasar systems holds the key to resolving one of the most important scientific questions: the Hubble constant ($H_{0}$) tension. The commonly used Markov chain Monte Carlo (MCMC) method has been too time-consuming to achieve this goal, yet recent work has shown that convolution neural networks (CNNs) can be an alternative with seven orders of magnitude improvement in speed. With 31,200 simulated strongly lensed quasar images, we explore the usage of Vision Transformer (ViT) for simulated strong gravitational lensing for the first time. We show that ViT could reach competitive results compared with CNNs, and is specifically good at some lensing parameters, including the most important mass-related parameters such as the center of lens $\theta_{1}$ and $\theta_{2}$, the ellipticities $e_1$ and $e_2$, and the radial power-law slope $\gamma'$. With this promising preliminary result, we believe the ViT (or attention-based) network architecture can be an important tool for strong lensing science for the next generation of surveys. The open source of our code and data is in \url{https://github.com/kuanweih/strong_lensing_vit_resnet}.

Aggretriever: A Simple Approach to Aggregate Textual Representation for Robust Dense Passage Retrieval

Pre-trained transformers has declared its success in many NLP tasks. One thread of work focuses on training bi-encoder models (i.e., dense retrievers) to effectively encode sentences or passages into single-vector dense vectors for efficient approximate nearest neighbor (ANN) search. However, recent work has demonstrated that transformers pre-trained with mask language modeling (MLM) are not capable of effectively aggregating text information into a single dense vector due to task-mismatch between pre-training and fine-tuning. Therefore, computationally expensive techniques have been adopted to train dense retrievers, such as large batch size, knowledge distillation or post pre-training. In this work, we present a simple approach to effectively aggregate textual representation from the pre-trained transformer into a dense vector. Extensive experiments show that our approach improves the robustness of the single-vector approach under both in-domain and zero-shot evaluations without any computationally expensive training techniques. Our work demonstrates that MLM pre-trained transformers can be used to effectively encode text information into a single-vector for dense retrieval. Code are available at: https://github.com/castorini/dhr