Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Waterloo

$41,000.00

Multi-Agent Communication and Collaboration for Specialized Large Language Models

FY 2024-25

Summary
Context. The rapid proliferation of large language models (LLMs) has predominantly focused on general-purpose systems. But as the need for personalized and privacy-preserving LLMs grows, and with hardware becoming more accessible, a future brimming with diverse LLMs, each with their unique expertise, is on the horizon. This flourishing landscape of LLMs offers a unique opportunity for models to leverage each other's distinct knowledge, thereby reducing redundancy, conserving resources, and addressing privacy concerns. For example, consider the healthcare sector, where different LLMs are trained as medical specialists, such as oncologists, cardiologists, and geriatricians. A cardiology-focused LLM could consult a geriatrics-specialized LLM when faced with senior citizens and provide a more comprehensive diagnosis. Objectives. Building on this vision, this proposal investigates multi-agent communication and collaboration among LLMs, aiming to harness the potential that arises from their specialization. The research is centered on four primary objectives: O1. Induce a Latent Language for Inter-LLM Communication: Creating a shared language among models, facilitating knowledge exchange. O2. Enhance the Interpretability of the Latent Language for LLMs: Making model interactions transparent and understandable to humans, promoting human oversight. O3. Enable Communication for Models across Modalities: Bridging the gap between models that specialize in different data types, from text to images, enabling multi-modal interactions. O4. Foster a Division of Labor among Specialized Models: Encouraging models to specialize in specific domains, promoting a division of labor rather than universal capability. Impact. This research program will lead to a paradigm shift from singular, monolithic AI models to a collaborative network of specialized models. Such a shift promises a democratized AI landscape, paving the way for smaller entities to train and use their own expert models. Furthermore, this research will provide pivotal training for the next generation of AI professionals and equip them with a comprehensive skill set, from data handling to model training. Therefore, the proposed program will prepare the students for a myriad of roles in the ever-evolving AI landscape and ensure Canada remains at the vanguard of global AI advancements. In summary, this proposal envisions an AI future where models, akin to human teams, collaborate, communicate, and collectively address challenges, transcending the limitations of individual expertise.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2024-05178
Name nom
Deng, Yuntian
Department département
David R. Cheriton School of Computer Science
Organization id
33
Institution établissement
University of Waterloo
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
41000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2801
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Natural language and speech understanding
Research subject fr
Langage naturel et reconnaissance de la parole
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Multi-Agent Communication and Collaboration for Specialized Large Language Models
Keywords
large language models; language model agents; multi-agent systems; latent variable modeling; interpretability; multi-modality models; structured prediction; knowledge offloading; specialized models; division of labor
Application summary
Context. The rapid proliferation of large language models (LLMs) has predominantly focused on general-purpose systems. But as the need for personalized and privacy-preserving LLMs grows, and with hardware becoming more accessible, a future brimming with diverse LLMs, each with their unique expertise, is on the horizon. This flourishing landscape of LLMs offers a unique opportunity for models to leverage each other's distinct knowledge, thereby reducing redundancy, conserving resources, and addressing privacy concerns. For example, consider the healthcare sector, where different LLMs are trained as medical specialists, such as oncologists, cardiologists, and geriatricians. A cardiology-focused LLM could consult a geriatrics-specialized LLM when faced with senior citizens and provide a more comprehensive diagnosis. Objectives. Building on this vision, this proposal investigates multi-agent communication and collaboration among LLMs, aiming to harness the potential that arises from their specialization. The research is centered on four primary objectives: O1. Induce a Latent Language for Inter-LLM Communication: Creating a shared language among models, facilitating knowledge exchange. O2. Enhance the Interpretability of the Latent Language for LLMs: Making model interactions transparent and understandable to humans, promoting human oversight. O3. Enable Communication for Models across Modalities: Bridging the gap between models that specialize in different data types, from text to images, enabling multi-modal interactions. O4. Foster a Division of Labor among Specialized Models: Encouraging models to specialize in specific domains, promoting a division of labor rather than universal capability. Impact. This research program will lead to a paradigm shift from singular, monolithic AI models to a collaborative network of specialized models. Such a shift promises a democratized AI landscape, paving the way for smaller entities to train and use their own expert models. Furthermore, this research will provide pivotal training for the next generation of AI professionals and equip them with a comprehensive skill set, from data handling to model training. Therefore, the proposed program will prepare the students for a myriad of roles in the ever-evolving AI landscape and ensure Canada remains at the vanguard of global AI advancements. In summary, this proposal envisions an AI future where models, akin to human teams, collaborate, communicate, and collectively address challenges, transcending the limitations of individual expertise.