# HJYao00/Mulberry

[NIPS'25 Spotlight] Mulberry, an o1-like Reasoning and Reflection MLLM Implemented via Collective MCTS

Repository: https://github.com/HJYao00/Mulberry
Canonical: https://ross.abutalabs.com/products/mulberry
Language: Python
License Family: other
Last push: 2026-01-16T08:02:56+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 62, release rhythm 35, longevity 44
- inputs: {"age_days": 618, "days_push": 229, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1243, forks 113 (observed 2026-08-28T04:04:06.660358+00:00)

## What it is
Mulberry is a research implementation of an o1-like multimodal large language model (MLLM) that performs step-by-step reasoning and reflection via collective Monte Carlo Tree Search (MCTS). It includes released model checkpoints (LLaVA/Llama and Qwen2-VL based), reasoning inference code, and evaluation code accompanying a NeurIPS 2025 spotlight paper.

## Use cases
- run o1-style reasoning with a multimodal LLM
- generate reasoning and reflection traces with MCTS
- evaluate a reasoning MLLM on multimodal benchmarks
- fine-tune a multimodal model on reasoning SFT data
- reproduce a NeurIPS 2025 MLLM reasoning paper

## When to choose
- you need o1-like chain-of-thought reasoning for image+text inputs
- you want to experiment with MCTS-based reasoning data generation
- you want ready checkpoints and evaluation code for reasoning MLLMs

## When to avoid
- you need a production-ready inference server with support guarantees
- you need a permissively licensed project (no license is specified)
- you only need text-only reasoning without multimodal inputs

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, llm-training, agent-framework
- domain: large-language-models, machine-learning, computer-vision, artificial-intelligence
- platform: python
- tags: multimodal, mcts, reasoning, reflection, mllm, chain-of-thought, research-code, neurips, research, gpu, linux

## Member repositories
- HJYao00/Mulberry (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:06.660358+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:08:17.662424+00:00, confidence not recorded.
  - readme: https://github.com/HJYao00/Mulberry (fetched 2026-08-28T04:04:06.660358+00:00, sha c3c459830df8)
- Data as of 2026-08-30T08:39:29.467469+00:00.