> ## Documentation Index
> Fetch the complete documentation index at: https://staging.docs.trywhitney.com/llms.txt
> Use this file to discover all available pages before exploring further.

# SDPO (Self-Distillation Policy Optimization)

> Teacher = the same base model, conditioned on feedback about the student's own attempt.

**Contract:** teacher = the same base model, conditioned on feedback about
the student's own prior attempt (not a full reference answer — that's
[OPSD](/distillation/opsd)).

**Caller math:** Same teacher-forced mechanism as OPSD — both use
`advantage = teacher_logprob - student_logprob` — following
["Reinforcement Learning via Self-Distillation"](https://arxiv.org/abs/2601.20802)
(Hübotter et al., 2026): "the current model conditioned on feedback as a
self-teacher... without any external teacher or explicit reward model."

<Note>
  Distinct from the unrelated, same-acronym "sDPO: Don't Use Your Data All at Once" — sequential preference-data
  subsetting for offline DPO, not self-distillation.
</Note>

**Whitney mapping:** Same as OPSD — a [frozen second Whitney run](/distillation) —
with the teacher prompt built from feedback or critique instead of a golden
answer.

**Fail closed:** You must supply a real tokenized feedback-conditioned
prompt — there is no default critique.

```python theme={null}
def feedback_prompt_tokens() -> list[int]:
    """Your own tokenized question + feedback/critique prompt."""
    ...


def build_loss(student_sample, teacher_logprobs, cycle, request, ctx):
    del teacher_logprobs  # recomputed below against the feedback-conditioned prompt instead
    prompt_tokens = request["prompt"]["token_ids"]
    teacher_prompt = feedback_prompt_tokens()
    data = []
    for sequence in student_sample["sequences"]:
        completion_tokens = sequence["tokens"]
        full_tokens = teacher_prompt + completion_tokens
        result = ctx.client.sampler_operation(
            ctx.teacher_run_id, ctx.teacher_sampler_id, "logprobs", {"input": {"token_ids": full_tokens}}
        )
        teacher_seq_logprobs = [float(v) for v in result["logprobs"][-len(completion_tokens):]]
        student_seq_logprobs = sequence["logprobs"]
        per_token_advantage = [
            teacher_lp - student_lp
            for student_lp, teacher_lp in zip(student_seq_logprobs, teacher_seq_logprobs, strict=True)
        ]
        data.append(rl_datum(prompt_tokens, completion_tokens, student_seq_logprobs, per_token_advantage))
    return forward_backward_payload(data, "importance_sampling")
```

`rl_datum` and `forward_backward_payload` are the shared [datum helpers](/cookbooks/helpers#datum-helpers).