Researcher Questions scPerturb Efficacy as Primary Scaling Axis for Cell Models
anshulkundaje · x · 2026-08-06
Stanford professor Anshul Kundaje expresses confusion over why many efforts are heavily betting on scPerturb data as the primary scaling modality for single-cell models.
He notes there is little evidence that scaling well in these model cell systems provides much advantage to in vivo cell states. He argues that without a time component and with arbitrary endpoints, the bang-for-the-buck of scaling scPerturb might be lower than other data axes. However, he clarifies the data remains super useful for analyses involving decent proxies like certain immune cell systems.
Related event: Stanford Scholar Questions Efficacy of Single-Cell Perturbation Data for AI(2 posts)→
More from Research
- NBER Paper: 19% of Workers Retroactively Edit Profiles, AI Skills Surge — steverathje2 · 2026-08-06
- Nature Publishes Expanded Codebook of Human Transcription Factor DNA-Binding Specificity — anshulkundaje · 2026-08-06
- Reddit User Discovers New LLM Attack Vector: Non-Instructional Text Prefix Bypasses RLHF — Historical-Cod-2537 · 2026-08-06
- Hundred-Page Language Models Book: build LLMs from scratch with PyTorch — burkov · 2026-08-06
- Dark Hundred-Page Language Models Book released, hands-on LLM guide — burkov · 2026-08-06
- Researcher Analyzes Kimi K2.5: Multi-Agent Communication May Induce RL Reward Hacking — soldni · 2026-08-06