MalTree: Multi-modal Malware Embeddings

Multi-modal embeddings for 103,883 malware samples across 538 families, released with the paper "MalTree: Tracing Malware Evolution from Embeddings at Scale" (ICML 2026). Each file is a JSON object keyed by the SHA256 of the sample binary. Raw samples are not distributed; the original binaries may be retrievable from those hashes via VirusTotal. See the accompanying code repository for the schema and usage.