Skip to content

Radar ·

Three approaches to on-policy self-distillation

Published on February 12, 2026, the article examines OPSD, SDFT, and SDPO, on-policy self-distillation methods using denser feedback, as well as benchmarks and limitations.

Published on February 12, 2026, the article compares three on-policy self-distillation methods: OPSD, SDFT, and SDPO. According to the description, they use denser feedback; the article also covers benchmarks and limitations. The comparison is presented as useful to engineers studying how these methods use feedback to improve model reasoning.

To consult and verify the claims, read the original article and check its descriptions of each method, benchmarks, and stated limitations. The supplied summary gives no specific benchmark results, so it does not support attributing any particular advantage or performance to a method.

Original source: Três abordagens de self-distillation on-policy

Curation · AgentLog

Get the essentials

Privacy, AI engineering and security in your inbox.

At most one digest per day, when there is news. Cancel at any time. We use your account's verified email.

Privacy

Looking for something else?DocumentationContact