Official Implementation of "Reasoning Language Models: A Blueprint"

This is the repo of developing reasoning models in the specific domain of financial, aim to enhance models capabilities in handling financial reasoning tasks.

🔥🔥🔥Breaking long thought processes of o1-like LLMs, such as DeepSeek-R1, QwQ

Reasoning-from-Zero using gemma.JAX.nnx on TPUs

📖Curated list about reasoning abilitiy of MLLM, including OpenAI o1, OpenAI o3-mini, and Slow-Thinking.

This repository contains the implementation of our research on optimizing Retrieval-Augmented Generation (RAG) systems for technical domains. Our work addresses the unique challenges of precise information extraction from complex, domain-specific documents by introducing token-aware evaluation metrics and synthetic data generation pipeline.

A curated list of awesome open-source and open-weight language models or methods focused on reasoning capabilities.

Official code for "Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning", ICLR 2025.

Replication study of DeepSeek-R1. Explores pure RL without SFT for post-training for reasoning capability, leveraging (1) veRL framework, (2) Knights and Knaves (K&K) logic puzzle dataset, and (3) small-scale base model.

📖Curated list about reasoning abilitiy of MLLM, including OpenAI o1, OpenAI o3-mini, and Slow-Thinking.