Carbon-X training mix revealed: vertebrates dominate, human genomes favored in post-training
LoubnaBenAllal1 · x · 2026-10-22
The Carbon-X author published the species distribution of its midtraining corpus: within the eukaryotic genomic subset, other vertebrates account for 43%, mammals 28%, invertebrates 19%, plants 7%, with fungi and protozoa minimal. Post-training data skews toward human genomes, so the model performs best there.
More from Research
- isCORAL: PKU team does in silico lineage tracing from transcriptome snapshots alone — arjunrajlab · 2026-10-22
- IBM's Zlatko Minev publishes 76 quantum computing lectures, 3,098 slides, all free — CatAstro_Piyush · 2026-10-22
- Tsinghua's Design Anything uses intent-execution graphs to keep design agents on-brief at scale — Xuchao Ren · 2026-10-22
- DeepMind researchers map the coming agent economy where AIs trade beyond human oversight — weballergy · 2026-10-22
- Wav2Vec2 pretrained on 5 years of dolphin recordings beats general bioacoustic models — emaros96 · 2026-10-22
- OpenWhistle: largest open dolphin-vocalization dataset (180k whistles, 114h) is a NeurIPS Spotlight — emaros96 · 2026-10-22