# microsoft/SoM

[arXiv 2023] Set-of-Mark Prompting for GPT-4V and LMMs

Repository: https://github.com/microsoft/SoM
Canonical: https://ross.abutalabs.com/products/som
Language: Python
License: MIT
License Family: permissive
Last push: 2024-08-19T07:51:11+00:00

## Health v2 (maintenance only)
Score: 18/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 8, longevity 75
- inputs: {"age_days": 1052, "days_push": 744, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1557, forks 111 (observed 2026-08-28T04:05:03.384666+00:00)

## What it is
A research toolbox from Microsoft implementing Set-of-Mark (SoM) visual prompting, which overlays numbered spatial marks on images to improve visual grounding in GPT-4V and other large multimodal models. It includes mark-generation code using segmentation models, a GPT-4V demo, and an evaluation benchmark.

## Use cases
- overlay numbered marks on images to prompt GPT-4V
- improve visual grounding of multimodal LLMs
- build GUI or web navigation agents with visual prompting
- evaluate GPT-4V on visual reasoning benchmarks
- generate set-of-mark prompts from segmentation models

## When to choose
- you want to enhance GPT-4V or other LMMs with spatial grounding on images
- you are building agents that navigate UIs or detect anomalies via multimodal models
- you are reproducing or extending the SoM research

## When to avoid
- you need a production-ready end-user application rather than research code
- you do not use large multimodal models or OpenAI APIs
- you need actively maintained tooling with long-term support

## Facets
- artifact type: library
- maturity: maintenance
- function: computer-vision, image-processing, prompt-engineering, machine-learning
- domain: computer-vision, large-language-models, artificial-intelligence
- platform: python, cross-platform
- tags: visual-prompting, gpt-4v, multimodal-llm, set-of-mark, image-segmentation, research

## Member repositories
- microsoft/SoM (main) score 18

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:03.384666+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 2026-08-30T04:29:57.384160+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/SoM (fetched 2026-08-28T04:05:03.384666+00:00, sha ceeccc0cb633)
- Data as of 2026-08-30T08:39:29.467469+00:00.