kalomaze: You don't need an analytic transfer theory, just a learnable transfer-extrapolation function
kalomaze · x · 2026-09-12
Continuing the RLVR generalization discussion, kalomaze argues you don't necessarily need a strong mechanistic theory of transfer (which may not exist in closed form) before training. What you need instead is a strong learned function of transfer extrapolation for a given policy — something he says is constructible, replacing crude heuristics like domain similarity.
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