Glossary · Term

credit-assignment SFT

← all terms

Definition

Salvaging failed AI training trajectories by training only on the parts where the agent was actually making progress.

A supervised fine-tuning method that uses an LLM to estimate per-step value along failed teacher trajectories and trains the student only on segments preceding the critical mistake.

Mentioned in 1 episode

  1. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI