maitrix-org/llm-reasoners

Health v2 · maintenance only

Adoption not part of the score

2343 stars · 204 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

LLM Reasoners is a Python library that provides a standardized modular framework for advanced LLM reasoning, implementing algorithms like MCTS (RAP), Tree-of-Thoughts, Guided Search, and GRACE with pluggable search algorithms, reward functions, and world models. It includes an interactive reasoning-tree visualizer, AutoRace reasoning-chain evaluation, and integrates with high-performance backends such as SGLang, HuggingFace Transformers, and OpenAI API.

Use cases

• implement tree-of-thoughts reasoning with LLMs
• run MCTS-based planning for LLM problem solving
• compare different LLM reasoning algorithms on benchmarks
• visualize and debug LLM reasoning trees
• evaluate reasoning chains automatically with AutoRace
• scale inference-time compute with process reward models
• build custom reward functions and world models for LLM search

When to choose

• you need state-of-the-art reasoning/search algorithms (ToT, RAP, MCTS) for LLMs
• you want a modular framework to prototype new reasoning methods
• you need to visualize or evaluate multi-step reasoning chains
• you want efficient reasoning with SGLang-backed structured generation

When to avoid

• you only need simple single-shot prompting or chat completions
• you need a production chatbot or RAG pipeline rather than reasoning research
• you work outside Python or need a no-code tool
• you require fine-tuning or training of LLMs rather than inference-time reasoning

Facets

library · maturity active

machine-learning llm-inference agent-framework prompt-engineering data-visualization large-language-models artificial-intelligence machine-learning python cross-platform llm-reasoning tree-of-thoughts mcts chain-of-thought search-algorithms reward-functions world-model sglang reasoning-visualization auto-race natural-language-processing research

Member repositories

For agents

markdown · JSON · MCP: product_card(name="maitrix-org/llm-reasoners")