kalomaze: Weak Transfer Persists Because Labs Just Stratify Domains and Pray
kalomaze · x · 2026-09-12
In a thread with StefanGliga, kalomaze argues capability transfer remains weak because the industry isn't explicitly targeting it via better transfer prediction—current practice is overwhelmingly 'stratify across explicit domains and pray.' He adds that humans have bad intuitions about what helps broadly, so labs scale synthetic versions of what SWEs already care about.
More from Research
- Neural Thickets Revisited: Frozen-Model Logit Perturbations Match Full-Weight Perturbation Ensembles — kastnerkyle · 2026-10-31
- CS312 course post explains sharpness and the edge of stability in LLM pretraining — tatsu_hashimoto · 2026-10-31
- AI decodes wolf howls: 500K hours of Yellowstone audio analyzed by models — aza · 2026-10-31
- NBER paper: training your AI replacement can actually raise workers' pay — joshgans · 2026-10-31
- Trata's Hedge-Bench: no model tops 50% on real-world financial forecasting — ycombinator · 2026-10-31
- Anthropic's internal Claude model claims 2^o(n) 3-SAT algorithm, purportedly refuting the ETH — ctjlewis · 2026-10-31