Abstract: The present disclosure relates to techniques for enhancing retrieval-augmented generation (RAG) pipelines by efficiently reducing a length of retrieved context into compressed representations. The length of the compressed representations may be orders of magnitude smaller than the length of the retrieved context and thus achieve a substantial speed-up in response generation by a RAG, while maintaining similar accuracy as compared to traditional RAG. In some aspects, pretty simple compression (PISCO) techniques are disclosed for training and deploying generative artificial intelligence (AI) models within a RAG pipeline. Accordingly, the PISCO techniques may employ two generative AI models: a compressor model producing the compressed representations and a generator model generating a response based on a received user query and the compressed representations of the retrieved passages.

Abstract: A context pruning system retrieves a passage initially selected and ranked by an initial passage selection model, such as a model that determines similarity based on cosine similarities between the query and various passages. The context pruning system re-ranks the passages using a re-ranking model that may also be used to prune clauses from the passages. Alternatively, the context pruning system may prune clauses from the passages using a separate clause selection model. Clauses preserved from the pruning may be used to prompt a natural language processor for a context-informed result, and the context-informed result may trigger a downstream action such as display of an interface that uses the context-informed result or operation of a robot that uses the context-informed result to physically move or audibly speak.

Abstract: Information retrieval methods employ a neural network encoder configured to receive a dense representation and generate a composite code comprising C clusters of dimension L from the dense representation. An activation function is configured to generate a sparse composite code from the composite code. The sparse composite code comprises a binary representation. An index can be generated using the sparse composite code.

Abstract: A neural model for representing an input sequence over a vocabulary in a ranker of a neural information retrieval model. An input sequence is embedded based at least on the vocabulary. An importance of each token over the vocabulary is predicted with respect to each token of the embedded input sequence. A predicted term importance of the input sequence over the vocabulary is determined by performing an activation over the embedded input sequence.

Abstract: Information retrieval methods employ a neural network encoder configured to receive a dense representation and generate a composite code comprising C clusters of dimension L from the dense representation. An activation function is configured to generate a sparse composite code from the composite code. The sparse composite code comprises a binary representation. An index can be generated using the sparse composite code.

Abstract: A system and method perform cross-modal information retrieval, by generating a graph representing the set of media objects. Each node of the graph corresponds to a media object and is labeled with a set of features corresponding to a text part of the respective media object. Each edge between two nodes represents a similarity between a media part of the two nodes. A first relevance score is computed for each media object of the set of media objects that corresponds to a text-based score. A second relevance score is computed for each media object by inputting the graph into a graph neural network. The first relevance score and the second relevance score are combined to obtain a final ranking score for each media object.

Abstract: A system and method perform cross-modal information retrieval, by generating a graph representing the set of media objects. Each node of the graph corresponds to a media object and is labeled with a set of features corresponding to a text part of the respective media object. Each edge between two nodes represents a similarity between a media part of the two nodes. A first relevance score is computed for each media object of the set of media objects that corresponds to a text-based score. A second relevance score is computed for each media object by inputting the graph into a graph neural network. The first relevance score and the second relevance score are combined to obtain a final ranking score for each media object.