Kundaje: model perturbation effects in latent space to dodge noisy single-cell data
anshulkundaje · x · 2026-10-10
In an exchange with Fabian Theis, Anshul Kundaje adds that phenotype-level prediction can happen in latent space, which is far less noisy than observed gene expression. The reply is part of their ongoing discussion about better-defined training objectives for perturbation-prediction models.
Related event: Theis and Kundaje Debate Phenotype-First Perturbation Modeling(4 posts)→
More from Research
- Bacteria sense collisions with neighbors to self-regulate collective movement, Nature Microbiology — NikoMcCarty · 2026-10-10
- Inference Auctions: bidding for scarce LLM serving capacity without breaking KV-cache gains — nhaghtal · 2026-10-10
- ThunderSyncRL speeds up synchronous agentic RL by up to 1.9x with zero policy staleness — StanfordAILab · 2026-10-10
- Prompt injection detection benchmarks have badly mislabeled data, researcher finds — evilsocket · 2026-10-10
- Researcher Replicates Claude 4's 'Bliss Attractor' Eval: LLMs Fall Into Distinct Obsession Basins — rgblong · 2026-10-10
- YC-backed team builds 8,000 sqft wet-lab in 30 days, AI compounds kill $13B pest — ycombinator · 2026-10-10