Official codebase for the Atari-GPT paper, benchmarking multimodal LLMs as low-level policies in Atari.

Official codebase for the Rating-Based Reinforcement Learning paper presented at AAAI 2024.

An implementation enabling Llama 3.2 3B to play ASCII Breakout, optimized for Apple Silicon using the MLX framework.

Implementation allowing various Large Language Models (Google Gemini Pro/Flash 2.0, OpenAI GPT-4o, Llama 3.2 3B) to act as agents playing an ASCII version of the Atari game Breakout.

A nanoGPT model (based on Karpathy's GPT-2 implementation) trained entirely from scratch using PPO to play ASCII Breakout.

An AI-powered tool utilizing the Gemini API to provide instant, constructive feedback on research papers, aiding researchers with actionable insights and writing suggestions.

Simplified Python implementations of Rating-based Reinforcement Learning (RbRL) and Preference-based Reinforcement Learning (PbRL), designed for clarity using common RL libraries.

A straightforward Python implementation of the Monte Carlo Tree Search (MCTS) algorithm.

A simple chatbot application with a graphical user interface (GUI) leveraging the Gemini API for generating responses.

Sign Language Recognition model developed using Recurrent Neural Networks (GRU, LSTM, Simple RNN).

Exploration and implementation focused on Adversarial Multi-Agent Reinforcement Learning (MARL) concepts.

A stock prediction application utilizing an LSTM (Long Short-Term Memory) model, featuring a Tkinter GUI for user interaction.