Selective retrieval improves retrieval-augmented generation (RAG) by reducing distractions from low-quality retrievals and improving efficiency. However, existing approaches under-utilize the inherent knowledge of large language models (LLMs), leading to suboptimal retrieval decisions and degraded generation performance. To bridge this gap, we propose Self-Routing RAG (SR-RAG), a novel framework that binds selective retrieval with knowledge verbalization. SR-RAG enables an LLM to dynamically decide between external retrieval and verbalizing its own parametric knowledge. To this end, we design a multi-task objective that jointly optimizes an LLM on knowledge source selection, knowledge verbalization, and response generation. We further introduce dynamic knowledge source inference via nearest neighbor search to improve the accuracy of knowledge source decision under domain shifts. Fine-tuning three LLMs with SR-RAG significantly improves both their response accuracy and inference latency. Compared to the strongest selective retrieval baseline, SR-RAG reduces retrievals by 29% while improving the performance by 5.1%.

Hallucinations and unfaithful synthesis due to inaccurate prompts with insufficient semantic details are widely observed in multimodal generative models. A prevalent strategy to align multiple modalities is to fine-tune the generator with a large number of annotated text-image pairs. However, such a procedure is labor-consuming and resource-draining. The key question we ask is: can we enhance the quality and faithfulness of text-driven generative models beyond extensive text-image pair annotations? To address this question, we propose Knowledge Pursuit Prompting (KPP), a zero-shot framework that iteratively incorporates external knowledge to help generators produce reliable visual content. Instead of training generators to handle generic prompts, KPP employs a recursive knowledge query process to gather informative external facts from the knowledge base, instructs a language model to compress the acquired knowledge for prompt refinement, and utilizes text-driven generators for visual synthesis. The entire process is zero-shot, without accessing the architectures and parameters of generative models. We evaluate the framework across multiple text-driven generative tasks (image, 3D rendering, and video) on datasets of different domains. We further demonstrate the extensibility and adaptability of KPP through varying foundation model bases and instructions. Our results show that KPP is capable of generating faithful and semantically rich content across diverse visual domains, offering a promising solution to improve multimodal generative models.

Domain-specific language understanding requires integrating multiple pieces of relevant contextual information. For example, we see both suicide and depression-related behavior (multiple contexts) in the text ``I have a gun and feel pretty bad about my life, and it wouldn't be the worst thing if I didn't wake up tomorrow''. Domain specificity in self-attention architectures is handled by fine-tuning on excerpts from relevant domain specific resources (datasets and external knowledge - medical textbook chapters on mental health diagnosis related to suicide and depression). We propose a modified self-attention architecture Knowledge-infused Self Attention Transformer (KSAT) that achieves the integration of multiple domain-specific contexts through the use of external knowledge sources. KSAT introduces knowledge-guided biases in dedicated self-attention layers for each knowledge source to accomplish this. In addition, KSAT provides mechanics for controlling the trade-off between learning from data and learning from knowledge. Our quantitative and qualitative evaluations show that (1) the KSAT architecture provides novel human-understandable ways to precisely measure and visualize the contributions of the infused domain contexts, and (2) KSAT performs competitively with other knowledge-infused baselines and significantly outperforms baselines that use fine-tuning for domain-specific tasks.

Accurately reporting what objects are depicted in an image is largely a solved problem in automatic caption generation. The next big challenge on the way to truly humanlike captioning is being able to incorporate the context of the image and related real world knowledge. We tackle this challenge by creating an end-to-end caption generation system that makes extensive use of image-specific encyclopedic data. Our approach includes a novel way of using image location to identify relevant open-domain facts in an external knowledge base, with their subsequent integration into the captioning pipeline at both the encoding and decoding stages. Our system is trained and tested on a new dataset with naturally produced knowledge-rich captions, and achieves significant improvements over multiple baselines. We empirically demonstrate that our approach is effective for generating contextualized captions with encyclopedic knowledge that is both factually accurate and relevant to the image.

Contextualized entity representations learned by state-of-the-art deep learning models (BERT, GPT, T5, etc) leverage the attention mechanism to learn the data context. However, these models are still blind to leverage the knowledge context present in the knowledge graph. Knowledge context can be understood as semantics about entities, and their relationship with neighboring entities in knowledge graphs. We propose a novel and effective technique to infuse knowledge context from knowledge graphs for conceptual and ambiguous entities into models based on transformer architecture. Our novel technique project knowledge graph embedding in the homogeneous vector-space, introduces new token-types for entities, align entity position ids, and a selective attention mechanism. We take BERT as a baseline model and implement "KnowledgeInfused BERT" by infusing knowledge context from ConceptNet and WordNet, which significantly outperforms BERT over a wide range of NLP tasks over eight different GLUE datasets. KI-BERT-base model even outperforms BERT-large for domain-specific tasks like SciTail and academic subsets of QQP, QNLI, and MNLI.

With the emerging branch of incorporating factual knowledge into pre-trained language models such as BERT, most existing models consider shallow, static, and separately pre-trained entity embeddings, which limits the performance gains of these models. Few works explore the potential of deep contextualized knowledge representation when injecting knowledge. In this paper, we propose the Contextualized Language and Knowledge Embedding (CoLAKE), which jointly learns contextualized representation for both language and knowledge with the extended MLM objective. Instead of injecting only entity embeddings, CoLAKE extracts the knowledge context of an entity from large-scale knowledge bases. To handle the heterogeneity of knowledge context and language context, we integrate them in a unified data structure, word-knowledge graph (WK graph). CoLAKE is pre-trained on large-scale WK graphs with the modified Transformer encoder. We conduct experiments on knowledge-driven tasks, knowledge probing tasks, and language understanding tasks. Experimental results show that CoLAKE outperforms previous counterparts on most of the tasks. Besides, CoLAKE achieves surprisingly high performance on our synthetic task called word-knowledge graph completion, which shows the superiority of simultaneously contextualizing language and knowledge representation.

Several recent efforts have been devoted to enhancing pre-trained language models (PLMs) by utilizing extra heterogeneous knowledge in knowledge graphs (KGs), and achieved consistent improvements on various knowledge-driven NLP tasks. However, most of these knowledge-enhanced PLMs embed static sub-graphs of KGs ("knowledge context"), regardless of that the knowledge required by PLMs may change dynamically according to specific text ("textual context"). In this paper, we propose a novel framework named DKPLM to dynamically select and embed knowledge context according to textual context for PLMs, which can avoid the effect of redundant and ambiguous knowledge in KGs that cannot match the input text. Our experimental results show that DKPLM outperforms various baselines on typical knowledge-driven NLP tasks, indicating the effectiveness of utilizing dynamic knowledge context for language understanding. Besides the performance improvements, the dynamically selected knowledge in DKPLM can describe the semantics of text-related knowledge in a more interpretable form than the conventional PLMs. Our source code and datasets will be available to provide more details for DKPLM.