docs/usermanuals/en/operations/troubleshooting.md: broken example — Format auto-detection wrong because the first JSONL row differs; fix with format: "preference". (DOC-D3B04B-011) #229

Wave 1 · theme D1-10: Documentation-only fixes · highest severity High · 1 finding(s) · effort docs fix

Summary

The suggested fix makes the config fail to load.

Primary finding

DOC-D3B04B-011 — docs/usermanuals/en/operations/troubleshooting.md: broken example — Format auto-detection wrong because the first JSONL row differs; fix with format: "preference".

severity High · already tracked: KNOWN-OPEN C31-DOC-SCHEMA-PROSE docs/roadmap/phase-16-trust-surface-hardening.md:785-786

Where. docs/usermanuals/en/operations/troubleshooting.md:118-122 (Turkish mirror: docs/usermanuals/tr/operations/troubleshooting.md:118-122) · type: broken example

The document says. Format auto-detection wrong because the first JSONL row differs; fix with format: "preference".

What the code does. There is no format key at the top level or under data (both fail validation). Format is advisory and inferred from the dataset's column names. The trainer is chosen by training.trainer_type, and _validate_trainer_columns enforces the required columns (forgelm/data.py:23-48, 712-720).

Code references. forgelm/data.py:23-48, forgelm/data.py:712-720

User impact. The suggested fix makes the config fail to load.

Proposed correction. Cause: rows have inconsistent keys, or training.trainer_type does not match the columns. Fix: set training.trainer_type explicitly (e.g. dpo for chosen/rejected) and make every row carry the same columns.

Turkish mirror. the Turkish page has the same error; fix both in the same PR

Acceptance criteria

• Each statement is re-checked against the code at fix time.
• Documentation listed above is corrected in English and in its Turkish mirror in the same PR.
• Every finding in this issue is resolved or explicitly closed with a reason.

_{Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.}