# UMass-Embodied-AGI/3D-LLM

Code for  3D-LLM: Injecting the 3D World into Large Language Models

Repository: https://github.com/UMass-Embodied-AGI/3D-LLM
Canonical: https://ross.abutalabs.com/products/3d-llm
Language: Python
License: MIT
License Family: permissive
Last push: 2024-06-06T19:14:00+00:00

## Health v2 (maintenance only)
Score: 28/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 81
- inputs: {"age_days": 1136, "days_push": 818, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1212, forks 77 (observed 2026-08-28T04:04:00.309042+00:00)

## What it is
3D-LLM is the research code for a large language model that takes 3D representations (objects and scenes) as input, built on BLIP-2/LAVIS. It includes pretrained and finetuned checkpoints for 3D visual question answering tasks like ScanQA, SQA3d, and 3DMV_VQA.

## Use cases
- run inference on 3D objects and scenes with a language model
- finetune an LLM on 3D visual question answering datasets
- answer questions about 3D scenes like ScanQA or SQA3d
- experiment with injecting 3D features into large language models
- reproduce NeurIPS 2023 3D-LLM paper results

## When to choose
- you need an LLM that accepts 3D scene or object representations
- you want to benchmark on 3D VQA datasets like ScanQA or SQA3d
- you are researching multimodal 3D-language models

## When to avoid
- you need a production-ready or maintained product with active updates
- you want a simple pip-installable library or Hugging Face integration
- you work with 2D images only

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-inference, llm-training, machine-learning, deep-learning
- domain: large-language-models, artificial-intelligence, computer-vision
- platform: python
- tags: 3d-representations, multimodal, visual-question-answering, research-code, neurips-2023, point-clouds, blip2, natural-language-processing, gpu, linux

## Member repositories
- UMass-Embodied-AGI/3D-LLM (main) score 28

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:00.309042+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 2026-08-30T06:17:58.787903+00:00, confidence not recorded.
  - readme: https://github.com/UMass-Embodied-AGI/3D-LLM (fetched 2026-08-28T04:04:00.309042+00:00, sha 9b19474aeeb0)
- Data as of 2026-08-30T08:39:29.467469+00:00.