Multi-agent learning & decision-making
My established work on credit assignment, communication and coordination: how agents learn and make decisions together.
ELLIS PHD STUDENT · UNIVERSITY OF MANCHESTER
Aditya
Kapoor.
I study learning and decision-making in intelligent systems.
My work spans multi-agent reinforcement learning and language-model-based planning. I am expanding into probabilistic and generative modelling, including diffusion models and large language models.
I enjoy finding mathematical connections across fields and following questions into unfamiliar territory. I want to understand how learning systems work, develop useful methods, and learn from people with different perspectives.
Manchester, UK
RESEARCH
Research interests & current directions.
Probabilistic & generative modelling
A developing direction in my research, exploring diffusion models, geometry and probabilistic approaches to learning.
Language models & agent systems
Work spanning multi-agent planning with language models and emerging investigations into structured decoding.
SELECTED PUBLICATIONS
Methods, agents and environments.
Peer-reviewed work. Asterisks denote equal contribution.
Assigning Credit with Partial Reward Decoupling in Multi-Agent Proximal Policy Optimization
Learns which teammates are relevant to an agent's updates to improve credit assignment in cooperative MARL.
Long-Horizon Planning for Multi-Agent Robots in Partially Observable Environments
LLaMAR: language-model-based multi-agent planning with a plan-act-correct-verify loop and evaluation under partial observability.
Learning Cooperative Multi-Agent Policies with Partial Reward Decoupling
Uses learned relevance between agents and rewards to reduce interference in cooperative policy learning.
SocNavGym: A Reinforcement Learning Gym for Social Navigation
A simulation environment for developing and evaluating socially aware robot navigation policies.
Low-Rank Agent-Specific Adaptation (LoRASA) for Multi-Agent Policy Learning
Explores low-rank agent-specific adaptation for multi-agent policy learning.
PREPRINTS
Recent manuscripts.
Public manuscripts; listed separately from peer-reviewed publications.
Learning what to say and how precisely: Efficient Communication via Differentiable Discrete Communication Learning
Studies end-to-end learning of discrete messages and their precision for communication-constrained multi-agent learning.
Redistributing Rewards Across Time and Agents for Multi-Agent Reinforcement Learning
Studies reward redistribution across the temporal and agent dimensions in cooperative MARL.
IN PROGRESS
Questions I am exploring now.
Ongoing investigations, with results and scope still being evaluated.
Communication precision in multi-agent reinforcement learning
Investigating stochastic quantisation and dithering for discrete inter-agent messages. Controlled comparisons examine message precision, communication cost and coordination performance against fixed-precision baselines.
Measuring contributions across agents and time
Developing controlled diagnostic studies of credit assignment in cooperative MARL, distinguishing predictive contribution scores from evidence about how interventions change team outcomes.
Structured decoding for masked language models
Studying structured and randomised decoding schedules for frozen masked language models, separating schedule structure from the effect of learning mixture weights at matched model-call budgets.
Diffusion models on curved spaces
Exploring how to fine-tune diffusion models on curved spaces, including rotations. I am studying how geometry shapes the modelling and learning problem; this work is still developing.
WIDENING THE LENS
Learning through research.
I am curious about ideas across mathematics, the sciences and beyond, and I enjoy research that leads me into unfamiliar territory. Information theory, information geometry, topology, neuroscience, cognitive science and psychology are a few examples of subjects I would like to explore.
My interests keep evolving as I encounter new questions and perspectives. I hope to deepen my understanding through research, conversations and collaboration, and I am always open to learning something unexpected.
My newer research directions include probabilistic generative modelling, structured decoding for masked language models, and fine-tuning diffusion models on curved spaces. These are developing parts of my research programme.
Working on a related question, or see an interesting connection? I am open to collaborations — send me an email.
More publications, preprints and patent applications
MAP-THOR: Benchmarking Long-Horizon Multi-Agent Planning Frameworks in Partially Observable Environments
Agent-Temporal Credit Assignment for Optimal Policy Preservation in Sparse Multi-Agent Reinforcement Learning
Concept-based Anomaly Detection in Retail Stores for Automatic Correction using Mobile Robots
Challenges in Applying Robotics to Retail Store Management
Auto-TransRL: Autonomous Composition of Vision Pipelines for Robotic Perception
Supervised category learning: When do participants use a partially diagnostic feature?
A toolkit to generate social navigation datasets
Effect of a colour-based descriptor and stimuli presentation mode in unsupervised categorization
DeepClean: Integrated Distortion Identification and Algorithm Selection for Rectifying Image Corruptions
Systems and methods for anomaly detection and correction
Methods and systems for autonomous task composition of vision pipelines using an algorithm selection framework
ABOUT
Background & collaboration.
I am pursuing a PhD in Computer Science at the University of Manchester through the ELLIS programme (November 2024–December 2028, expected).
My supervisory team is Omar Rivasplata (primary), Samuel Kaski (secondary), and Jan Peters (secondary and ELLIS co-supervisor at TU Darmstadt).
Previously, I was a predoctoral research fellow at Tata Consultancy Services Research & Innovation, working on multi-agent learning and robotics. I was mentored by Harshad Khadilkar, Mayank Baranwal and Jayavardhana Gubbi, and collaborated with Vighnesh Vatsal, Vartika Sengar and Nijil George.
I completed my B.E. (Hons.) in Computer Science and Information Systems at BITS Pilani, Goa. My undergraduate thesis on Partial Reward Decoupling (PRD) was carried out with Benjamin Freed, Howie Choset and Jeff Schneider at Carnegie Mellon University.
Alongside research, I spend time mentoring and building community projects, and exploring ways to make mathematics more inviting.
MENTORING & BUILDING
Learning with other people.
Open source & research mentoring
I have participated in Google Summer of Code (GSoC) as both a student and a mentor. I was also a founding member of CORD.ai, a student organisation where I mentored students and worked with them on research projects.
Some of those mentoring relationships developed into research collaborations, including papers with Sushant Swamy and Yash Bhisikar. I value helping students find questions they care about and grow into collaborators.
Project Kratos
As an undergraduate at BITS Pilani, Goa, I was part of the founding team of Project Kratos, our student Mars rover project. I worked on autonomous navigation and later helped mentor the navigation team.
COMMUNITY & EDUCATION
Projects I care about.
StraysCue
I run StraysCue, a nonprofit initiative that helps rescue stray animals in urban India, including Mumbai.
Agla Station
We recently started Agla Station to help students in Classes 9 and 10 in rural India discover educational and career possibilities. The aim is to make a wider range of paths visible while students are beginning to think about their next steps.
Mathematics through play
I am developing ways to teach mathematics through making, movement and performance — including origami, magic, Rubik’s cubes and dance.
I want children to encounter science as something they can explore: fold, move, try, get surprised, and ask why. My ambition is to make that curiosity part of learning from an early age, while helping students have fun and acquire new skills.
BEYOND RESEARCH
Books, movement
& curiosity.
Sport is a big part of my life. I have played football in the Goa League Division 2 and currently play futsal for the University of Manchester.
I also enjoy surfskating, calisthenics, gymnastics, bouldering, badminton and hiking. Away from sport, you might find me folding origami or playing chess.
I enjoy reading books and am exploring philosophy and politics to enrich my understanding of the world. Reading gives me space to encounter different perspectives, question my assumptions and think more carefully about how people and societies make sense of things.
There is an unfinished Welsh 3K hiking challenge on my list, too — one I am keen to return to.