# HumanMLLM/R1-Omni

Repository: https://github.com/HumanMLLM/R1-Omni
Canonical: https://ross.abutalabs.com/products/r1-omni
Language: Python
License Family: other
Last push: 2025-03-24T03:54:33+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 13, release rhythm 35, longevity 38
- inputs: {"age_days": 542, "days_push": 527, "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 1022, forks 73 (observed 2026-08-28T04:03:15.964258+00:00)

## What it is
R1-Omni is a research project applying Reinforcement Learning with Verifiable Reward (RLVR) to an omni-multimodal large language model for explainable emotion recognition from video and audio. It releases model weights (HumanOmni-0.5B variants) and training/inference code built on the HumanOmni base model.

## Use cases
- recognize emotions from video with audio and visual cues
- train a multimodal LLM with reinforcement learning and verifiable rewards
- get explainable reasoning for emotion recognition predictions
- fine-tune an omni-modal model on emotion datasets like MAFW and DFEW
- evaluate emotion recognition generalization on out-of-distribution data
- run inference with a 0.5B multimodal emotion model

## When to choose
- you need multimodal (audio+visual) emotion recognition with interpretable reasoning
- you want to experiment with RLVR training on multimodal LLMs
- you need open model weights for emotion recognition research

## When to avoid
- you need a production-ready, well-documented pipeline (setup and reproduction docs are incomplete)
- you need single-modality (video-only or audio-only) inference, which is not yet supported
- you require a permissive license - no license is specified

## Facets
- artifact type: library
- maturity: experimental
- function: machine-learning, deep-learning, llm-training, speech-recognition, video-processing, nlp
- domain: artificial-intelligence, large-language-models, machine-learning
- platform: python
- tags: emotion-recognition, multimodal, reinforcement-learning, rlvr, omni-model, research, audio, video, gpu, linux

## Member repositories
- HumanMLLM/R1-Omni (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:15.964258+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 2026-08-30T07:08:41.423944+00:00, confidence not recorded.
  - readme: https://github.com/HumanMLLM/R1-Omni (fetched 2026-08-28T04:03:15.964258+00:00, sha a01f68f19aa2)
- Data as of 2026-08-30T08:39:29.467469+00:00.