README.md

metadata

title: SGP-Tribe3
emoji: 🧠
colorFrom: green
colorTo: yellow
sdk: docker
pinned: false
license: cc-by-nc-4.0

SGP-Tribe3

Sentient Generative Principal — Brain Encoding Calibration System

A fork of TRIBE v2 (Meta AI) extended with:

• Full multimodal inference: video + audio + text simultaneously
• SGP 9-node parcellation using the Schaefer-200 cortical atlas
• Dual-stream (dorsal/ventral) activation separation
• White matter tract co-activation edge weight output
• Stimulus calibration matrix accumulation across sessions

API Endpoints

POST /predict

Upload a video file (mp4, max 120s). Returns SGP node activation profile.

Request: multipart/form-data with video field

Response:

{
  "stimulus_id": "uuid",
  "sgp_nodes": {
    "G1_broca": 0.73,
    "G2_wernicke": 0.61,
    "G3_tpj": 0.55,
    "G4_pfc": 0.48,
    "G5_dmn": 0.32,
    "G6_limbic": 0.67,
    "G7_sensory": 0.81,
    "G8_atl": 0.59,
    "G9_premotor": 0.44
  },
  "streams": {
    "ventral_mean": 0.64,
    "dorsal_mean": 0.58
  },
  "dominant_hemisphere": "left",
  "n_timesteps": 42,
  "n_vertices": 20484
}

GET /health

Returns model load status.

POST /warmup

Triggers model loading (call once after cold start).

Architecture

Built on the SGP-LLM framework. Nodes map to:

• G1 Broca — phonological production, syntactic processing
• G2 Wernicke — auditory comprehension, lexical-semantic decoding
• G3 TPJ — stream convergence, sensorimotor interface
• G4 PFC — executive control, coherence/veto
• G5 DMN — generativity, self-referential processing
• G6 Limbic — emotional weighting, memory consolidation
• G7 Sensory — primary perceptual input encoding
• G8 ATL — cross-modal semantic integration hub
• G9 Premotor — action planning, motor speech preparation

Citation

Built on TRIBE v2:

@article{dAscoli2026TribeV2,
  title={A foundation model of vision, audition, and language for in-silico neuroscience},
  author={d'Ascoli, Stéphane and Rapin, Jérémy and Benchetrit, Yohann and others},
  year={2026}
}