Open Source Deep Research Alternative to Reason and Search on Private Data. Written in Python.

MiniMax-M1, the world's first open-weight, large-scale hybrid-attention reasoning model.

R-KV: Redundancy-aware KV Cache Compression for Reasoning Models

MedReason: Eliciting Factual Medical Reasoning Steps in LLMs via Knowledge Graphs

Official implementation of the paper "Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space"

Simple extension on vLLM to help you speed up reasoning model without training.

xVerify: Efficient Answer Verifier for Reasoning Model Evaluations

Pivotal Token Search

Lightweight replication study of DeepSeek-R1-Zero. Interesting findings include "No Aha Moment", "Longer CoT ≠ Accuracy", and "Language Mixing in Instruct Models".

VeriThinker: Learning to Verify Makes Reasoning Model Efficient

Pure RL to post-train base models for social reasoning capabilities. Lightweight replication of DeepSeek-R1-Zero with Social IQa dataset.

[arXiv 2025] Can MLLMs Guide Me Home? A Benchmark Study on Fine-Grained Visual Reasoning from Transit Maps

Code for the 2025 ACL publication "Fine-Tuning on Diverse Reasoning Chains Drives Within-Inference CoT Refinement in LLMs"

R1-Code-Interpreter: Training LLMs to Reason with Code via Supervised and Reinforcement Learning

Agentic Deep Graph Reasoning Implementation

AI Lawyer is an intelligent reasoning legal assistant powered by DeepSeek , Ollama RAG and LangChain, designed to streamline legal research and document analysis. By leveraging retrieval-augmented generation (RAG), it provides precise legal insights, and contract summarization. With an intuitive Streamlit-based UI, analyze legal documents.

LLM finetuning for Sudoku solving

Turn stories, strategies, or systems into insight. Auto-generate Dialectical Wheels (DWs) from any text to reveal blind spots, surface polarities, and trace dynamic paths toward synthesis. DWs are semantic maps that expose tension, transformation, and coherence within a system—whether narrative, ethical, organizational, or technological.

Sudoku4LLM is a Sudoku dataset generator for training and evaluating reasoning in Large Language Models (LLMs). It offers customizable puzzles, difficulty levels, and 11 serialization formats to support structured data reasoning and Chain of Thought (CoT) experiments.

MANBench: Is Your Multimodal Model Smarter than Human?