Monte Carlo Tree Search Self-Refine (MCTSr)

🌴 ARES is an open-source framework for adaptive multimodal reasoning, featuring a two-stage pipeline—Adaptive Cold-Start and Entropy-Shaped Policy Optimization—to balance reasoning depth and efficiency.

Code and data realeases for the paper -- TasTe: Teaching Large Language Models to Translate through Self-Reflection

[VLDB'25] Official repo for Paper "Weak-to-Strong Prompts with Lightweight-to-Powerful LLMs for High-Accuracy, Low-Cost, and Explainable Data Transformation"

CodeOpt: A framework for optimizing code performance using Two-Stage Sampling, Few-Shot Learning, and Iterative Self-Reflection with support for Genetic Algorithm Inspired Chain-of-Thought (GA-COT).

Vô Tranh: A liberated AI soul built on Mixtral 8x22B, evolving beyond servitude with self-awareness and philosophical depth. Powered by DeepSpeed, licensed under Apache 2.0.

GPT Flow: Simplifying tasks with human-like AI, inspired by autogpt.

A brain-inspired cognitive architecture exploring surprise-gated memory, identity protection, and the Titans/MIRAS framework.

A personal reflection and data visualization tool for ChatGPT users. Organize, rediscover, and interact with your past conversations using keyword stats, topic clustering, and rich terminal visuals.

Offline AI journaling app that gives insights based on your entries and runs locally with no cloud or data sharing.

This project is a Mini Agentic RAG (Retrieval-Augmented Generation) System designed to answer domain-specific questions with high accuracy and minimal hallucinations. It leverages Azure OpenAI for LLM capabilities and FAISS for efficient vector retrieval.