WIP: P27.2 — TripleVerbalizer: KG triples → diverse training pairs via 4 grammar archetypes #320

Draft
toasterson wants to merge 2 commits from claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers into main
Owner

Anima work item 019f90cc-b77c-75c1-bcc6-1e45a650d684.

Second slice of the re-scoped Phase 27. Independent of Phase 26 — can run in parallel with P27.1.

Deliverable: convert verified KG triples into natural-language training pairs using the four grammar archetypes from the phase plan, producing varied surface forms rather than one template per triple.

Why diversity is the requirement, not a nicety: fine-tuning on a single phrasing per fact teaches the model the template, not the fact. The archetypes exist to prevent that, so "does it produce genuinely varied output" IS the acceptance test, not a stretch goal.

Acceptance:

  • Each archetype is exercised and produces grammatical output on a fixed triple sample.
  • Measurable surface diversity across archetypes for the same triple (state the metric used).
  • Only VERIFIED triples are eligible — unverified/low-confidence KG content must not leak into training data. Enforce it here rather than trusting the caller.
  • Deterministic for a fixed triple + archetype + seed, so training runs are reproducible.

Feeds: P27.3 (the LoRA trainer consumes these pairs).

⚠ akh-medu repo. Ignore inherited crates/anima-ai/... paths from the original WI-11 triage.

Anima work item `019f90cc-b77c-75c1-bcc6-1e45a650d684`. Second slice of the re-scoped Phase 27. **Independent of Phase 26** — can run in parallel with P27.1. **Deliverable:** convert verified KG triples into natural-language training pairs using the four grammar archetypes from the phase plan, producing varied surface forms rather than one template per triple. **Why diversity is the requirement, not a nicety:** fine-tuning on a single phrasing per fact teaches the model the template, not the fact. The archetypes exist to prevent that, so "does it produce genuinely varied output" IS the acceptance test, not a stretch goal. **Acceptance:** - Each archetype is exercised and produces grammatical output on a fixed triple sample. - Measurable surface diversity across archetypes for the same triple (state the metric used). - Only VERIFIED triples are eligible — unverified/low-confidence KG content must not leak into training data. Enforce it here rather than trusting the caller. - Deterministic for a fixed triple + archetype + seed, so training runs are reproducible. **Feeds:** P27.3 (the LoRA trainer consumes these pairs). ⚠ akh-medu repo. Ignore inherited `crates/anima-ai/...` paths from the original WI-11 triage.
P27.2 — TripleVerbalizer: KG triples → diverse training pairs via 4 grammar archetypes
Some checks failed
CI / check-seshat (pull_request) Failing after 2h0m18s
CI / test-akh-medu (pull_request) Failing after 2h0m19s
CI / publish-chart (pull_request) Failing after 2h0m19s
CI / docker-seshd (pull_request) Failing after 2h4m51s
CI / check-akh-medu (pull_request) Failing after 2h5m5s
eval-smoke / eval-smoke (pull_request) Failing after 2h5m13s
95bf1092c1
Anima implementation session.
P27.2 — TripleVerbalizer: KG triples → diverse training pairs via 4 grammar archetypes
Some checks failed
CI / check-akh-medu (pull_request) Failing after 2h3m53s
eval-smoke / eval-smoke (pull_request) Failing after 2h3m52s
CI / docker-seshd (pull_request) Failing after 2h3m55s
CI / check-seshat (pull_request) Failing after 2h4m1s
CI / test-akh-medu (pull_request) Failing after 2h4m1s
CI / publish-chart (pull_request) Failing after 2h4m3s
fad3fa379b
Anima implementation session.
toasterson force-pushed claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers from fad3fa379b
Some checks failed
CI / check-akh-medu (pull_request) Failing after 2h3m53s
eval-smoke / eval-smoke (pull_request) Failing after 2h3m52s
CI / docker-seshd (pull_request) Failing after 2h3m55s
CI / check-seshat (pull_request) Failing after 2h4m1s
CI / test-akh-medu (pull_request) Failing after 2h4m1s
CI / publish-chart (pull_request) Failing after 2h4m3s
to 4b97133652
Some checks failed
CI / check-akh-medu (pull_request) Has been cancelled
CI / check-seshat (pull_request) Has been cancelled
CI / publish-chart (pull_request) Has been cancelled
CI / test-akh-medu (pull_request) Has been cancelled
CI / docker-seshd (pull_request) Has been cancelled
eval-smoke / eval-smoke (pull_request) Has been cancelled
2026-08-19 18:30:47 +00:00
Compare
toasterson force-pushed claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers from 4b97133652
Some checks failed
CI / check-akh-medu (pull_request) Has been cancelled
CI / check-seshat (pull_request) Has been cancelled
CI / publish-chart (pull_request) Has been cancelled
CI / test-akh-medu (pull_request) Has been cancelled
CI / docker-seshd (pull_request) Has been cancelled
eval-smoke / eval-smoke (pull_request) Has been cancelled
to 1e5833be7f
Some checks failed
CI / publish-chart (pull_request) Has been cancelled
CI / check-akh-medu (pull_request) Has been cancelled
CI / check-seshat (pull_request) Has been cancelled
CI / docker-seshd (pull_request) Has been cancelled
CI / test-akh-medu (pull_request) Has been cancelled
eval-smoke / eval-smoke (pull_request) Has been cancelled
2026-08-19 20:01:02 +00:00
Compare
toasterson force-pushed claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers from 1e5833be7f
Some checks failed
CI / publish-chart (pull_request) Has been cancelled
CI / check-akh-medu (pull_request) Has been cancelled
CI / check-seshat (pull_request) Has been cancelled
CI / docker-seshd (pull_request) Has been cancelled
CI / test-akh-medu (pull_request) Has been cancelled
eval-smoke / eval-smoke (pull_request) Has been cancelled
to 51d895d9bc
Some checks are pending
CI / check-seshat (pull_request) Waiting to run
CI / check-akh-medu (pull_request) Waiting to run
CI / test-akh-medu (pull_request) Waiting to run
CI / docker-seshd (pull_request) Waiting to run
CI / publish-chart (pull_request) Waiting to run
eval-smoke / eval-smoke (pull_request) Waiting to run
2026-08-19 21:30:49 +00:00
Compare
Some checks are pending
CI / check-seshat (pull_request) Waiting to run
CI / check-akh-medu (pull_request) Waiting to run
CI / test-akh-medu (pull_request) Waiting to run
CI / docker-seshd (pull_request) Waiting to run
CI / publish-chart (pull_request) Waiting to run
eval-smoke / eval-smoke (pull_request) Waiting to run
This pull request is marked as a work in progress.
This branch is out-of-date with the base branch
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers:claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers

Merge

Merge the changes and update on Forgejo.

Warning: The "Autodetect manual merge" setting is not enabled for this repository, you will have to mark this pull request as manually merged afterwards.

git switch main
git merge --no-ff claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git rebase main
git switch main
git merge --ff-only claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git rebase main
git switch main
git merge --no-ff claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch main
git merge --squash claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch main
git merge --ff-only claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git switch main
git merge claude/wi-019f90cc-p27-2-tripleverbalizer-kg-triples-divers
git push origin main
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
toasterson/akh-medu!320
No description provided.