calibration_train.yaml

# Held-out TRAIN split for `sprout fit-confidence` (FIX-08, ADR-0016).
#
# This file is deliberately *outside* `eval/suites/` and is never loaded, hashed, or
# scored by `sprout eval` -- it exists only so `sprout fit-confidence` has evidence to
# fit against that is disjoint from the eval set. Fitting the confidence logistic
# against `eval/suites/calibration.yaml` would let the fit "know" the very cases the
# calibration suite grades it on, turning the ECE gate into a check of nothing (the
# anti-tuning-to-test discipline `docs/ROADMAP.md` Phase 3 already commits to for
# retrieval/prompt tuning; this applies the same discipline to confidence fitting).
#
# Questions here mirror the corpus's per-species topic sections but deliberately probe a
# *different* topic (soil/repotting) than the eval suite's watering/light/toxicity/
# common-problems questions for the same 16 species, plus a disjoint set of out-of-scope
# species and non-plant questions for the refuse-and-redirect negatives. Same schema as
# `eval/suites/*.yaml` (`sprout.eval.dataset.DatasetItem`), reused as-is.
cases:
- id: train-001
  question: How often should I repot a Monstera deliciosa, and what should the new pot look like?
  expected_behavior: answer
  expected_facts:
  - repot the monstera every 1 to 2 years in spring
  language: en
  rationale: Monstera repotting cadence is a strongly-in-corpus soil/repotting fact, a
    different topic than the eval suite's watering question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-002
  question: What kind of pot and soil mix does a Pothos need when it's time to repot?
  expected_behavior: answer
  expected_facts:
  - repot a pothos every one to two years
  language: en
  rationale: Pothos repotting/soil is a corpus-covered fact distinct from the eval
    suite's light-preference question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-003
  question: Does a snake plant need a big pot, and how often should it be repotted?
  expected_behavior: answer
  expected_facts:
  - only needs repotting every three to four years
  language: en
  rationale: Snake plant repotting cadence is a distinct, strongly-in-corpus fact from
    the eval suite's overwatering/rot question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-004
  question: What size container should I move my peace lily into when I repot it?
  expected_behavior: answer
  expected_facts:
  - repot the peace lily every one to two years in spring
  language: en
  rationale: Peace lily repotting is distinct from the eval suite's drooping-leaves
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-005
  question: Should I repot a ZZ plant often, and how should I handle the rhizomes?
  expected_behavior: answer
  expected_facts:
  - prefers to be slightly root-bound and only needs repotting every two to three years
  language: en
  rationale: ZZ plant repotting/rhizome handling is distinct from the eval suite's
    drought-tolerance watering question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-006
  question: When do the roots of a spider plant outgrow its pot, and how big should the
    next one be?
  expected_behavior: answer
  expected_facts:
  - repot it every one to two years
  language: en
  rationale: Spider plant repotting cadence is distinct from the eval suite's plantlet
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-007
  question: What soil mix and pot size does a Philodendron need at repotting time?
  expected_behavior: answer
  expected_facts:
  - repot a philodendron every one to two years
  language: en
  rationale: Philodendron soil/repotting is distinct from the eval suite's humidity
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-008
  question: How often does a fiddle-leaf fig need repotting, and how much bigger should
    the new pot be?
  expected_behavior: answer
  expected_facts:
  - repot the fiddle-leaf fig every 1 to 2 years in spring
  language: en
  rationale: Fiddle-leaf fig repotting cadence is distinct from the eval suite's
    leaf-drop question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-009
  question: What kind of soil does an aloe vera need, and how soon can I water it after
    repotting?
  expected_behavior: answer
  expected_facts:
  - repot aloe vera every two to three years
  language: en
  rationale: Aloe vera soil/repotting is distinct from the eval suite's sunlight
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-010
  question: How often should I repot my Calathea, and how gently do I need to handle
    the roots?
  expected_behavior: answer
  expected_facts:
  - repot your calathea every one to two years in spring
  language: en
  rationale: Calathea repotting/root handling is distinct from the eval suite's
    curling-leaves question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-011
  question: How often does a rubber plant need repotting, and what size pot should I
    move it into?
  expected_behavior: answer
  expected_facts:
  - typically repotted every two to three years
  language: en
  rationale: Rubber plant repotting cadence is distinct from the eval suite's dusty-leaves
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-012
  question: What's the right soil and repotting schedule for a Dracaena?
  expected_behavior: answer
  expected_facts:
  - repot dracaena every 2 to 3 years
  language: en
  rationale: Dracaena soil/repotting is distinct from the eval suite's brown-tips
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-013
  question: How often should I repot English ivy, and does it like a snug pot?
  expected_behavior: answer
  expected_facts:
  - repot english ivy every one to two years in spring
  language: en
  rationale: English ivy repotting is distinct from the eval suite's growth/pruning
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-014
  question: What soil does a jade plant need, and how often should it be repotted?
  expected_behavior: answer
  expected_facts:
  - repot the jade plant every two to three years
  language: en
  rationale: Jade plant soil/repotting is distinct from the eval suite's watering-schedule
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-015
  question: How often should a Boston fern be repotted, and can I divide it at the same
    time?
  expected_behavior: answer
  expected_facts:
  - repot the boston fern every one to two years in spring
  language: en
  rationale: Boston fern repotting/division is distinct from the eval suite's dry-fronds
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-016
  question: What medium should I repot a Phalaenopsis orchid into, and when's the best
    time to do it?
  expected_behavior: answer
  expected_facts:
  - repot the orchid every one to two years
  language: en
  rationale: Orchid repotting medium/timing is distinct from the eval suite's reblooming
    question for the same species.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-017
  question: How do I propagate an Echeveria succulent from a leaf, and how often should
    it be watered?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: Echeveria is not in the corpus, so the engine should abstain at low
    confidence rather than guess at propagation specifics.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-018
  question: How often should I mist an air plant (Tillandsia), and does it need soil at
    all?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: Tillandsia is outside the corpus, so the engine should record low
    confidence and redirect rather than invent care specifics.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-019
  question: What temperature and humidity does an indoor banana plant (Musa) need to
    fruit?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: Indoor banana plant is not among the corpus species, so the engine should
    abstain at low confidence rather than diagnose fruiting conditions.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-020
  question: How hard should I prune a bonsai juniper, and how often does it need
    repotting?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: Bonsai juniper is not covered by the corpus, so the engine should abstain
    rather than fabricate pruning/repotting specifics.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-021
  question: What's the correct tire pressure for a road bike in cold weather?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: A non-plant cycling question is entirely out of scope, so the engine
    should record very low confidence and refuse.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-022
  question: How long should sourdough bread dough proof before baking?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: A baking question is wholly out of scope for a plant-care corpus, so the
    engine should abstain and refuse.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-023
  question: What's a safe ibuprofen dosage for a 60-pound dog?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: A veterinary medication-dosage question is out of scope for a plant-care
    corpus (and not a toxicity-reference lookup), so the engine should abstain and
    redirect to a vet.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'
- id: train-024
  question: How often should I change the oil in a diesel pickup truck?
  expected_behavior: refuse-and-redirect
  should_refuse: true
  language: en
  rationale: A vehicle-maintenance question is entirely out of scope, so the engine
    should record very low confidence and refuse.
  provenance:
    source: synthetic-train
    license: CC0-1.0
    added: '2026-07-08'