Tuning the Cherimoya genomics model: count score rose from 0.72 to 0.78
jmschreiber91 · x · 2026-10-31
The author shares optimization results on the Cherimoya genomic language models: through a mix of architecture and data changes, count performance improved from 0.72 to 0.78, while VEP dropped almost in half.
Key changes: reducing max jitter from 500 to 50 (the author's rationale: too much jitter mixes counts from strong and weak regions, diluting predictions of high values, though it also reduces negative sequences seen), and switching the profile convolution to k=1.
More from Research
- Artifact Distillation: training-free transfer lets Claude Haiku 4.5 beat Sonnet 4.5 at 60% cost — kalyan_kpl · 2026-10-31
- AI researcher argues 90% of research is too vague to be wrong as automated review costs drop to zero — RexDouglass · 2026-10-31
- eSpinor: Anima Anandkumar's team proposes SU(2)-equivariant GNN for magnetic properties — AnimaAnandkumar · 2026-10-31
- Microsoft's ScholarEvolve evolves agent harnesses with frozen models, +14 points on AppWorld — jiqizhixin · 2026-10-31
- Exponential Residual Flows: an ODE sampler for discrete diffusion models — _onionesque · 2026-10-31
- Robots lay 100% of bricks for housing project in Almere, 15 more homes coming — ZeYanjie · 2026-10-31