メニュー
TRIBE v2 tri-modal foundation model for in-situ fMRI brain-to-image decoding with synthetic data augmentation. Addresses low-data regime challenges in brain decoding.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/hiyenwong/ai_collection --skill tribev2-brain-foundation-modelコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
SOC 職業分類に基づく
| name | tribev2-brain-foundation-model |
| description | TRIBE v2 tri-modal foundation model for in-situ fMRI brain-to-image decoding with synthetic data augmentation. Addresses low-data regime challenges in brain decoding. |
| category | neuroscience |
| tags | ["brain decoding","fMRI","foundation model","data augmentation","multimodal","encoding model","video","audio","language"] |
| version | 1 |
| arxiv_id | 2606.06345v1 |
| authors | Yohann Benchetrit, Marlène Careil, Simon Dahan, Hubert Banville, Stéphane d'Ascoli |
| published | 2026-06-04T00:00:00.000Z |
| activation_keywords | ["brain decoding","fMRI","TRIBE","foundation model","data augmentation","brain-to-image","encoding model"] |
TRIBE v2 is a large encoding model pretrained on more than 1000 hours of fMRI responses to video, audio, and language stimuli. It enables synthetic data augmentation for boosting brain-to-image decoding in low-data regimes.
arXiv: 2606.06345v1
Key Contribution: Addresses the fundamental limitation in brain decoding — availability of labeled neural data — by using pretrained foundation models to generate synthetic fMRI data for augmentation.
Grid-based evaluation:
When augmentation helps:
When augmentation is less effective:
Input Stimuli → TRIBE v2 Encoder → Predicted fMRI → Quality Filter → Augmented Training Set
Quality filtering criteria:
Option A: Direct Augmentation
# Mix real and synthetic fMRI data
augmented_dataset = real_fMRI + synthetic_fMRI[quality > threshold]
model.train(augmented_dataset)
Option B: Pretrain-then-Finetune
# Use TRIBE v2 as pretrained backbone
pretrained_encoder = TRIBE_v2.load()
finetuned_decoder = pretrained_encoder.adapt(target_dataset)
Traditional limitation:
TRIBE v2 solution:
Problem: Reconstruct viewed images from brain activity TRIBE v2 approach: Augment with synthetic fMRI generated from image dataset
Problem: Train decoder on few subjects, generalize to new subjects Solution: Use TRIBE v2 pretrained representations as shared basis
Problem: Decode brain responses to stimuli not in training set Approach: Generate synthetic fMRI for new stimuli using TRIBE v2 encoder
def optimal_quality_threshold(real_data, synthetic_data):
"""
Find threshold maximizing augmentation benefit
"""
thresholds = np.linspace(0.5, 0.95, 20)
best_threshold = None
best_accuracy = 0
for thresh in thresholds:
filtered_synthetic = synthetic_data[quality > thresh]
augmented = real_data + filtered_synthetic
accuracy = evaluate_decoder(augmented)
if accuracy > best_accuracy:
best_accuracy = accuracy
best_threshold = thresh
return best_threshold, best_accuracy
When target dataset differs from TRIBE v2 pretraining distribution:
| Dataset Size | Recommended Synthetic Ratio | Quality Threshold |
|---|---|---|
| < 10 subjects | 50-70% | > 0.85 |
| 10-50 subjects | 30-50% | > 0.80 |
| 50-100 subjects | 20-30% | > 0.75 |
| > 100 subjects | 10-20% | > 0.70 |
Including all synthetic data: Low-quality predictions hurt more than help
Ignoring domain mismatch: TRIBE v2 trained on video/audio/language
Over-augmenting: Too much synthetic data overwhelms real signal
No validation of synthetic quality: Synthetic may not match real patterns
# Brain-to-image decoding with TRIBE v2 augmentation
class TRIBEv2AugmentedDecoder:
def __init__(self, pretrained_encoder_path):
self.encoder = load_tribe_v2(pretrained_encoder_path)
self.decoder = ImageDecoder()
def augment_dataset(self, real_fmri, stimuli_pool, quality_threshold=0.80):
"""
Generate synthetic fMRI for stimuli and filter by quality
"""
synthetic_fmri = []
for stimulus in stimuli_pool:
predicted_fmri = self.encoder.predict(stimulus)
quality = self.encoder.confidence(stimulus)
if quality > quality_threshold:
synthetic_fmri.append((predicted_fmri, stimulus))
# Mix real and synthetic
augmented = real_fmri + synthetic_fmri
return augmented
def train(self, real_fmri, stimuli_pool):
"""
Train decoder with augmented data
"""
augmented = self.augment_dataset(real_fmri, stimuli_pool)
self.decoder.fit(augmented)
return self.decoder
def decode(self, fmri_signal):
"""
Decode image from fMRI
"""
return self.decoder.predict(fmri_signal)
# Usage
decoder = TRIBEv2AugmentedDecoder('tribe_v2_pretrained.pt')
decoder.train(subject_fmri, image_dataset)
reconstructed_image = decoder.decode(new_fmri)
Activation: Use this skill when working on brain decoding, fMRI data augmentation, foundation models for neuroscience, or low-data regime brain-to-image reconstruction.