AVI Scores Are PHRED-Scaled and Traceable to Contributing Features
anshulkundaje · x · 2026-09-18
Thread segment from Anshul Kundaje: raw AVI scores from the trained classifier rank all 9 billion possible variants and are PHRED-scaled — AVI 10 = top 10%, 20 = top 1%, 30 = top 0.1%, 40 = top 0.01%. A cool feature: you can trace back through the model which features (binding, accessibility, splicing, expression) drive a high score.
More from Research
- Epoch launches Benchmark Reviews, auditing 15 AI benchmarks: only 4 verified, 9 flawed — stochasticchasm · 2026-09-18
- RL agents invent their own diagnostic renderings to ground code understanding, sparking RL scaling optimism — teortaxesTex · 2026-09-18
- Grounded SI: egocentric data at scale hinges on hand tracking in wild, long-tail scenarios — micoolcho · 2026-09-18
- GenBio AI co-founders publish 'A world model of the virtual cell' in Cell — HongyiWang10 · 2026-09-18
- Weeks after Navier-Stokes, Hodge Conjecture reportedly cracked: Millennium Problems falling fast — haider1 · 2026-09-18
- Meaning Spark Labs experiments with inference-time metacognitive scaffolding for LLMs — PeterBowdenLive · 2026-09-18