docs: soup train has no --task flag

`soup train --help` contains no `--task`; the task is a `soup.yaml` key. Two
places said otherwise, and both read as an instruction rather than as history:

- `docs/commands.md` listed `soup train --task unlearn` in the command
  cheat-sheet, where lines are copied straight into a terminal. Rewritten to
  `soup train  # task: unlearn`, the form the RAFT line directly below it
  already uses.
- `docs/training.md` said "`soup train --task unlearn` is live", two lines
  under a code block that correctly runs `soup train --config unlearn.yaml`.
  Now `task: unlearn`.

Checked every other `--task` in docs/: all 25 remaining are real flags on other
commands (`infer`, `data forge`, `bom emit`, `recipes search`, `eval coverage`,
`ship --task-eval/--task-mode`). Verified against the registered options, not
by reading a diff.

Left alone deliberately: CHANGELOG.md and benchmarks/gate-v0.72.4 also carry
the phrase, but they describe which code path broke rather than telling anyone
what to run, and the gate records are published verbatim on purpose. Four
docstrings in src/ and two in tests/ are narrative, not emitted to users.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

docs/commands.md

@@ -232,7 +232,7 @@ soup --no-audit-log <cmd> / SOUP_NO_AUDIT_LOG=1  Opt out of the per-command audi
 soup eval unlearning <run-id> --benchmark tofu|muse|wmdp  Forget Quality + Model Utility + PrivLeak verdict
 soup edit set --base <m> --method rome|memit|alphaedit|grace --subject "..." --target "..." [--output <dir>] [--device cpu] [--governor/--no-governor] [--registry-id <id>] [--cov-corpus <jsonl|txt>]  Live surgical knowledge edit (GPT-2 Conv1D + Llama; --cov-corpus = covariance-preconditioned ROME, rome-only; --plan-only available)
 soup edit diff <before-run> <after-run> --probes p.jsonl [--before-model <m> --after-model <m>]  Knowledge-injection diff (live before/after generation when both models given)
-soup train --task unlearn  NPO/SimNPO/RMU unlearning from data.forget_set (+ optional data.retain_set)
+soup train  # task: unlearn  NPO/SimNPO/RMU unlearning from data.forget_set (+ optional data.retain_set)
 soup train  # data.format='raft'  Answer-only span-mask RAFT training (golden+distractor docs, [doc-N] citations); generator-stage configs auto-link the latest RA-DIT retriever
 soup ra-dit --retriever-config <r.yaml> --generator-config <g.yaml> [--retriever-model <m>] [--plan-only]  One-shot two-stage RA-DIT: train retriever → record pairing → train generator
 soup eval citation <data> [--style bracket|inline|footnote] [--shuffle-seed N] [--output o.json]  Citation precision/recall/F1 over predictions or RAFT rows

docs/training.md

@@ -210,7 +210,7 @@ soup train --config unlearn.yaml --yes
 soup eval unlearning <run-id> --benchmark tofu --evidence evidence.json --output report.json
 ```
 
-`soup train --task unlearn` is live (v0.71.9): it loads a LoRA-wrapped policy, a frozen reference copy (NPO / RMU), and the forget / retain JSONL sets, then optimises the per-method loss — NPO's `(2/β)·mean(-logσ(-β·(π_logp − ref_logp)))` drives the policy's forget-set log-prob below the reference (= forgetting), while the retain set anchors capability. Run NPO/SimNPO **with** a `retain_set` — without one the policy has no utility anchor and Soup warns loudly.
+`task: unlearn` is live (v0.71.9): it loads a LoRA-wrapped policy, a frozen reference copy (NPO / RMU), and the forget / retain JSONL sets, then optimises the per-method loss — NPO's `(2/β)·mean(-logσ(-β·(π_logp − ref_logp)))` drives the policy's forget-set log-prob below the reference (= forgetting), while the retain set anchors capability. Run NPO/SimNPO **with** a `retain_set` — without one the policy has no utility anchor and Soup warns loudly.
 
 Three orthogonal axes: **Forget Quality** (pre/post forget-loss delta), **Model Utility** (retain-accuracy preserved), **PrivLeak** (membership-inference AUC distance from 0.5). Bundled mini-fixtures for all three benchmarks ship in the box (v0.71.1 added MUSE + WMDP alongside the existing TOFU set), so `--benchmark muse|wmdp` runs without supplying evidence. The WMDP forget-set probes ship **redacted** (placeholder prompts + `REFUSED` responses) — Soup never bundles verbatim hazardous-knowledge content.