Researcher: AlphaGenome distillation barely differs — variant correlations identical across models
jeffvierstra · x · 2026-09-19
Researcher jeffvierstra questions Google's AlphaGenome distillation, arguing that reporting 9 billion predictions isn't the goal — 'infinite GPUs/TPUs doesn't mean the predictions amount to much.' On closer inspection, signed Pearson correlations for variants are basically identical across all the models, suggesting the distillation mostly rescales noise; pre-distilled models aren't accessible to verify.
Related event: Benchmarks Question Distilled AlphaGenome's Performance(3 posts)→
More from Research
- Harvard Professors Admit AI Article Reviews Now Beat the Best Humans — iamfakhrealam · 2026-11-04
- Pre-pretraining gains traced to embedding init: N(0,0.02) alone lifts perplexity 9% — LauraRuis · 2026-11-04
- kalomaze shares loss curve of larger-backbone run on DeepSeek V3.1 base, eval loss down to ~0.82 — kalomaze · 2026-11-04
- Hobbyist rewrites Qwen 0.6B weights with 39.8M tokens, no gradients, in 3h12m — Altruistic_Heat_9531 · 2026-11-04
- SpecPilot agent boosts speculative decoding from 4.46× to 10.87×, no retraining — pmttyji · 2026-11-04
- Autonomous agent designs target-specific MSAs to improve AlphaFold 3 complex prediction — SJTU · 2026-11-04