Benchmark notes for RNA foundation models
How we compare embeddings, preserve biological signal, and detect benchmark leakage.
Working memory
Research notes
A public notebook for model design, dataset choices, evaluation failures, reading maps, and independent-research operating lessons. Notes can be provisional without pretending to be papers.
Planned series
Blog-like, but accountable to the science.
Bulk-to-single-cell translation logs
Design notes, failure cases, and sanity checks for cross-resolution transcriptomics.
Independent GPU science
Practical notes on learning with AI, training with constrained compute, and collaborating well.
Editorial rule
Every post should leave a reusable artifact.
A good note should include a question, evidence, limitations, and a next action. For technical posts, add code, config, dataset accession, model version, or a reproducibility note whenever possible.