Neyman Allocation

Uses

How it works

Evidence quality

Highlighted claims

• The allocation that minimizes gradient estimator variance under stratified sampling is the Neyman allocation, where phases are sampled proportionally to N_c times the square root of V_c. — Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking
• C_c, the success-failure action variance, serves as a computable lower-bound proxy for V_c that preserves the relative ordering of phases and requires no auxiliary model. — Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking
• Keep probabilities are updated every five steps using a minimum floor of 0.1 to prevent any phase from being permanently excluded due to a transient zero estimate. — Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking
• The realized chunk allocation during training closely matches the square-root-V_c-weighted prediction from Theorem 1, empirically confirming that C_c tracks V_c in practice. — Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking
• The cumulative C_c curve has a knee at approximately 20% of trajectory chunks, providing a principled heuristic for choosing the budget B in any new domain. — Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking