Repository navigation
[Epic] D1-10 · Documentation-only fixes (part 2/3) #858
Description
Epic · Wave 1 · 100 issue(s) · 100 finding(s) · part 2 of 3
Scope
Documentation-only fixes: each package is a single documentation finding. Re-verify every sentence against the code at fix time.
How to work on it
Every sub-issue below is independent and can be claimed on its own — see CONTRIBUTING.md, Finding Work: Issues and the Project Board. Progress shows in the sub-issue list and in the project's Sub-issues progress field.
Done when
- Every sub-issue is closed (fixed, or 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.
Activity
webhook.allow_private; 'Endpoint returns 4xx (silent in logs by default)'; 'Check audit_log.jsonl for web… (DOC-D3B04B-013) #231instructionsformat for SFT ={"prompt", "completion"}rows (DOC-D3B05-005) #232forgelm auditflags rows wherechosen == rejected('⚠ 12 chosen-rejected identical rows') (DOC-D3B05-007) #233ground_truthis passed to your reward functionreward(prompt, response, ground_truth); set `grpo_reward_model: "my_reward.rewa… (DOC-D3B05-008) #234data.dataset_name_or_path+data.extra_datasets(DOC-D3B05-009) #235forgelm auditcatches/flagschosen == rejectedcollisions and near-duplicates in preference data (DOC-D3B05-010) #236reference_free: true(less stable but skips the second model) (DOC-D3B05-011) #237severity_thresholds: {S1: 0.05, S2: 0.05, S5: 0.10, S10: 0.05}(DOC-D3B05-012) #238data.extra_datasets: ["data/sft.jsonl"]if you genuinely need to mix (DOC-D3B05-013) #23975 remaining items