segment-policy-optimization

Segment-level credit assignment for RL in LLMs using Monte Carlo advantage estimation, enabling precise reward attribution without critic models for improved reasoning task performance.

Independent GitHub snapshot · Automated classification · Original documentation governs