Research Interest

Reinforcement Learning + Large Language Models, Multi-Agent Learning, Artificial Intelligence for Social Good, Computational Game Theory

See Publication Page for More Information

AI for Social Good

We have been collaborating with a range of organizations to co-develop AI-driven solutions that address critical societal challenges. Selected projects and publications include:

• Mental Health and Public Behavioral Health. In collaboration with Collaborative Support Programs of New Jersey (CSPNJ)
  • PATIENT-Ψ: Using Large Language Models to Simulate Patients for Training Mental Health Professionals (2025) Best Paper Award at NeurIPS GenAI for Health Workshop 2024
• Media Monitoring for Environmental Conservation. In collaboration with the World Wild Fund for Nature (WWF).
  • NewsSerow (2024) Used by WWF Nepal and Colombia.
  • NewsPanda (2023) Used by WWF India and Nepal. Deployed Application Award at IAAI 2023.
• Improving Operational Efficiency in Food Rescue. In collaboration with and in use by 412 Food Rescue.
  •  Predicting and Presenting Task Difficulty for Crowdsourcing Food Rescue Platforms (2024) [Short Video]
  • A Recommender System for Crowdsourcing Food Rescue Platforms (2021)
  • Improving Efficiency of Volunteer-Based Food Rescue Operations (2020)
• Landmine Risk Prediction. Conducted a pilot field study in Colombia in partnership with a non-governmental humanitarian demining organization and an intergovernmental organization.
  • RELand: Risk Estimation of Landmines via Interpretable Invariant Risk Minimization (2023)
• Anti-Poaching Resource Allocation for Wildlife Conservation. Collaborative work with the World Wild Fund for Nature, Wildlife Conservation Society, Panthera, Rimba, Uganda Wildlife Authority
  • Improving Community-Participated Patrol for Anti-Poaching (2025)
  • Dual-mandate patrols: Multi-armed bandits for green security (2021) Best Paper Runner-Up, the 35th AAAI Conference on Artificial Intelligence (AAAI 2021).
  • Data-Driven Multimodal Patrol Planning for Anti-poaching (2021)
  • Using Game Theory in Real Time in the Real World: A Conservation Case Study (2019) Best Application System Demo (AAMAS 2019).
  • Exploiting data and human knowledge for predicting wildlife poaching (2018)
  • Taking It for a Test Drive: A Hybrid Spatio-Temporal Model for Wildlife Poaching Prediction Evaluated Through a Controlled Field Test (2017)
  • Cloudy with a Chance of Poaching: Adversary Behavior Modeling and Forecasting with Real-World Poaching Data (2017)
  • Deploying PAWS: Field Optimization of the Protection Assistant for Wildlife Security (2016) Innovative Application Award at IAAI 2016. Best Application of AI, Video Competition at AAAI 2016.

Reinforcement Learning + Large Language Models + Human-Centered AI

Over the past few years, we have been investigating how to integrate Large Language Models (LLMs) and Vision-Language Models (VLMs) with Reinforcement Learning (RL) to enhance AI’s ability to collaborate with and support humans or to strategically interact with humans in semi-cooperative settings. Selected projects and publications include:

• Improving LLM’s Reasoning Capabilities with Reinforcement Learning
  • Not All Rollouts are Useful: Down-Sampling Rollouts in LLM Reinforcement Learning (2025)
  • Cooperative Strategic Planning Enhances Reasoning Capabilities in Large Language Models (2024)
• Aligning/Improving Agents’ Decision Making through Human Feedback
  • M3HF: Multi-agent Reinforcement Learning from Multi-phase Human Feedback of Mixed Quality (2025)
  • Safe Multi-agent Reinforcement Learning with Natural Language Constraints (2024)
• Developing Strategic Language Agents
  • Grounded Persuasive Language Generation for Automated Marketing (2025)
  • Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game (2023)
• Interpretable Reinforcement Learning
  • LICORICE: Label-Efficient Concept-Based Interpretable Reinforcement Learning (2024)

Multi-Agent Learning

• Theoretical Analysis of Multi-Agent Learning
  • Multi-Agent Imitation Learning: Value is Easy, Regret is Hard (2024)
• Scalable Multi-Agent Learning
  • MESA: Cooperative Meta-Exploration in Multi-Agent Learning through Exploiting State-Action Space Structure (2024)
  • Evolutionary population curriculum for scaling multi-agent reinforcement learning (2020)
  • Robust multi-agent reinforcement learning via minimax deep deterministic policy gradient (2019)

Computational Game Theory

• Game Structure Learning
  • Deviate or Not: Learning Coalition Structures with Multiple-bit Observations in Games (2024)
  • Learning Coalition Structures with Games (2023)
  • What game are we playing? End-to-end learning in normal and extensive form games (2018) Distinguished Paper (IJCAI-ECAI 2018).
• Game Theory for Cyber Security
  • Teamwork Makes the Defense Work: Defense Resource Allocation with Composable Targets (2025)
  • Deceiving cyber adversaries: A game-theoretic approach (2018)
• Game Theory for Security
  • Multi-defender Security Games with Schedules (2023) Best Paper Award at The 14th Conference on Decision and Game Theory for Security (GameSec-23).
  • When Security Games Go Green: Designing Defender Strategies to Prevent Poaching and Illegal Fishing (2015) Outstanding Paper Award (IJCAI 2015, Computational Sustainability Track).

Other Topics

We also work on other topics related to human-centered AI and multi-agent systems. Many of these research projects are inspired by real-world problems. Selected projects and publications include:

• Mitigating Malicious Behavior in Peer Review
  • A One-Size-Fits-All Approach to Improving Randomness in Paper Assignment (2023) Deployed in NeurIPS 2024 [Blog Post]
  • Tradeoffs in Preventing Manipulation in Paper Bidding for Reviewer Assignment (2022) Outstanding Paper Award, ML Evaluation Standards workshop at ICLR 2022.
  • Near-Optimal Reviewer Splitting in Two-Phase Paper Reviewing and Conference Experiment Design (2021) Best Paper Honorable Mention at HCOMP 2022.
  • Mitigating Manipulation in Peer Review via Randomized Reviewer Assignments (2020) Deployed in AAAI 2022, AAAI 2023, KDD 2023, ICML 2024.
• AI for Transportation
  • Missing Pieces: How Framing Uncertainty Impacts Longitudinal Trust in AI Decision Aids — A Gig Driver Case Study (2024)
  • Purpose in the Machine: Do Traffic Simulators Produce Distributionally Equivalent Outcomes for Reinforcement Learning Applications? (2023)
• REALM: A Dataset of Real-World LLM Use Cases (2025)
• Global Rewards in Restless Multi-Armed Bandits (2024)