Tobin South
2024
Don’t forget private retrieval: distributed private similarity search for large language models
Guy Zyskind
|
Tobin South
|
Alex Pentland
Proceedings of the Fifth Workshop on Privacy in Natural Language Processing
While the flexible capabilities of large language models (LLMs) allow them to answer a range of queries based on existing learned knowledge, information retrieval to augment generation is an important tool to allow LLMs to answer questions on information not included in pre-training data. Such private information is increasingly being generated in a wide array of distributed contexts by organizations and individuals. Performing such information retrieval using neural embeddings of queries and documents always leaked information about queries and database content unless both were stored locally. We present Private Retrieval Augmented Generation (PRAG), an approach that uses multi-party computation (MPC) to securely transmit queries to a distributed set of servers containing a privately constructed database to return top-k and approximate top-k documents. This is a first-of-its-kind approach to dense information retrieval that ensures no server observes a client’s query or can see the database content. The approach introduces a novel MPC friendly protocol for inverted file approximate search (IVF) that allows for fast document search over distributed and private data in sublinear communication complexity. This work presents new avenues through which data for use in LLMs can be accessed and used without needing to centralize or forgo privacy.
2019
Podlab at SemEval-2019 Task 3: The Importance of Being Shallow
Andrew Nguyen
|
Tobin South
|
Nigel Bean
|
Jonathan Tuke
|
Lewis Mitchell
Proceedings of the 13th International Workshop on Semantic Evaluation
This paper describes our linear SVM system for emotion classification from conversational dialogue, entered in SemEval2019 Task 3. We used off-the-shelf tools coupled with feature engineering and parameter tuning to create a simple, interpretable, yet high-performing, classification model. Our system achieves a micro F1 score of 0.7357, which is 92% of the top score for the competition, demonstrating that “shallow” classification approaches can perform well when coupled with detailed fea- ture selection and statistical analysis.
Search
Co-authors
- Guy Zyskind 1
- Alex Pentland 1
- Andrew Nguyen 1
- Nigel Bean 1
- Jonathan Tuke 1
- show all...