ExoNet TESS High-Confidence Planet Candidate Catalog (arXiv:2604.15560) - Version 2.0

This catalog contains the high-confidence TESS exoplanet candidates identified by ExoNet, a calibrated multimodal deep learning framework combining 1D CNN, Multi-Head Attention, and late fusion of phase-folded light curves with stellar parameters. Model performance: Val AUC=0.9487, Test AUC=0.9549 Applied to: 4,720 unconfirmed TESS Planet Candidates (PC disposition) High-confidence (≥70%): 1,754 candidates Very high-confidence (≥85%): 1,098 candidates Habitable zone (200–400 K) + high-confidence: 52 candidates Earth-like habitable-zone candidates (Rp < 1.6 R⊕): 6 candidates Top candidate: TOI-6879.01 (96.84% confidence) Reference paper: Islam (2026), arXiv:2604.15560 v3 https://arxiv.org/abs/2604.15560