JevCodeDécouverte de la fonctionnalité Autoresearch
Exécute une boucle d'auto-recherche qui propose des questions TypeSafe, convertit le texte libre en caractéristiques numériques et utilise les erreurs du modèle pour améliorer un régresseur CatBoost supervisé.
Traduit automatiquement depuis le en, non relu. À utiliser comme référence rapide uniquement.
Les questions TypeSafe transforment le texte libre en caractéristiques numériques pour un modèle CatBoost supervisé ; utilisez une boucle de recherche automatique pour les découvrir.
CatBoost a besoin d’un tableau de nombres, et une note de dégustation n’en est pas un. Ce guide pratique construit le tableau à partir de questions sur la note, et aucune d’elles n’est rédigée à la main. Un LLM propose les questions, TypeSafe y répond pour chaque ligne, et CatBoost s’entraîne sur les réponses. La partie de recherche automatique correspond à ce qui suit : CatBoost indique quelles questions il a utilisées et quelles lignes il se trompe encore, l’appel de proposition suivant lit ce rapport, et la boucle se répète.
À la fin, vous disposez d’une boucle que vous pouvez appliquer à votre propre texte étiqueté, d’une courbe de l’erreur sur les données de validation par tour, et d’un tableau indiquant quelles questions le modèle final a le plus utilisées.
tasting note
|
v
38 TypeSafe answers
|-- 29 score questions x 2 columns = 58
| expected rubric level + answer uncertainty
`-- 9 noul questions x 1 column = 9
probability true
|
v
67 numeric columns --> CatBoost --> predicted critic score
held-out RMSE: 1.77 pointsUne réponse Score se décompose en deux colonnes : le niveau moyen que la réponse indique, et la dispersion autour de cette moyenne. Une réponse Noul est une seule probabilité, elle ne tient donc que dans une colonne.
Les données comprennent 2 000 critiques de vin : une note de dégustation en entrée, et la note du critique sur une échelle de 80 à 100 en sortie. L’erreur quadratique moyenne (RMSE) mesure l’erreur de prédiction en points de note critique, les écarts importants étant plus pénalisés, et une valeur plus basse étant meilleure. Chaque nombre du tableau ci-dessous provient des 800 critiques que ni le modèle ni la boucle n’ont jamais vues.
| comment la note devient un score | RMSE |
|---|---|
| prédire la note moyenne des lignes d’entraînement | 3.09 |
| le même CatBoost, en lisant la note comme des comptes de mots | 2.47 |
| demander à TypeSafe la note elle-même, recalée et décalée | 2.15 |
| 18 questions issues d’un appel de proposition, sans boucle | 1.87 |
| 38 questions après cinq tours de la boucle | 1.77 |
Les deux dernières lignes constituent la boucle. Un premier appel de proposition, sans aucune base encore, atteint 1,87. Quatre tours supplémentaires de lecture de ses pires prédictions abaissent ce chiffre à 1,77. La majeure partie du gain provient de cet premier appel, et l’apport des quatre tours suivants est mesuré plus bas.
Astuce — Vous souhaitez approfondir ce notebook ou l’appliquer à un autre problème ? Consultez Étapes suivantes.
from __future__ import annotations
import json
import os
import random
import textwrap
import urllib.request
from concurrent.futures import ThreadPoolExecutor
from pathlib import Path
from time import perf_counter
from typing import NamedTuple
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
from catboost import CatBoostRegressor
from cooksafe import JsonCache, make_playground_link
from IPython.display import Markdown, display
from typesafe_sdk import Noul, NoulCriteria, Score, TypeSafeClient
matplotlib.use("Agg") # headless render
TYPESAFE_MODEL = "jev-1.12"
FOLDS, REPEATS = 5, 3 # repeats steady the error at this sample size
CATBOOST = dict(
iterations=400,
depth=4,
learning_rate=0.05,
loss_function="RMSE",
verbose=0,
random_seed=0,
thread_count=1,
allow_writing_files=False,
)
client = TypeSafeClient(
# keyless kernels replay the cache
api_key=os.environ.get("TYPESAFE_API_KEY", "cache-only"),
base_url=os.environ.get("TYPESAFE_ENDPOINT"),
timeout=120.0,
)
json_cache = JsonCache(Path("json_cache.json"))
# ----------------------------------------------------------------- the specification
INTENSITY_LEVELS = [
"Not present in this note at all",
"Barely present - mentioned once, in passing",
"Present at a moderate level",
"Present strongly - the note dwells on it",
"Dominant - the note is largely about this",
]
PRESENCE_CRITERIA = NoulCriteria(
true="The note states this or clearly implies it",
false="The note gives no indication of this",
)
# Asking for the score outright: ten quality bands, rescaled onto the 80-100 critic scale.
SCORE_LEVELS = [
"Faulty or unpleasant - the note is mostly criticism",
"Barely acceptable - drinkable, with nothing to recommend it",
"Simple and sound - correct, plain, forgettable",
"Pleasant everyday wine - some appeal, little depth",
"Good - clear varietal character, well made",
"Very good - balanced, with something to say",
"Excellent - complex and structured",
"Outstanding - depth and length, built to age",
"Superb - among the best of its type",
"Profound - the note treats it as exceptional",
]
# Structured output requires every property in `required`, so unused fields come back empty.
PROPOSAL_SCHEMA = {
"type": "object",
"properties": {
"actions": {
"type": "array",
"items": {
"type": "object",
"properties": {
"op": {"type": "string", "enum": ["add", "revise", "drop"]},
"target": {"type": "string"},
"name": {"type": "string"},
"kind": {"type": "string", "enum": ["intensity", "presence"]},
"question": {"type": "string"},
},
"required": ["op", "target", "name", "kind", "question"],
"additionalProperties": False,
},
}
},
"required": ["actions"],
"additionalProperties": False,
}
PROPOSALS = 18 # actions the proposer may return per round
# The one string that knows this is about wine. Point it at your own label and text.
PROPOSER_TASK = f"""You are designing numeric features for a gradient-boosting model that
predicts the score a wine critic gave (an integer from 80 to 100) from the tasting note alone.
The model sees nothing but the features you design.
Return up to {PROPOSALS} actions. Each action is one of:
- {{"op": "add", "target": "", "name": ..., "kind": ..., "question": ...}}
A new feature.
- {{"op": "revise", "target": <name of an existing feature>, "name": ..., "kind": ...,
"question": ...}}
Replace that feature's question with better wording. Use this when a feature measures the
right thing badly: too narrow, too vague, or worded so nearly every note answers the same.
- {{"op": "drop", "target": <name of an existing feature>, "name": "", "kind": "intensity",
"question": ""}}
Remove a feature that is not earning its place.
`kind` is "intensity" for something with a degree, or "presence" for a yes/no fact.
`question` is what gets asked about one tasting note.
An "intensity" question is graded against this fixed five-level rubric, so word it so that the
levels make sense:
{chr(10).join(f" {i}. {level}" for i, level in enumerate(INTENSITY_LEVELS))}
A "presence" question is answered as the probability that it is true of the note.
Good features can be judged from the note's own words, vary from note to note, and carry
information about quality that the other features do not. Reviewers describe structure, fruit,
oak, length, complexity, and drinkability, and they also signal quality through word choice."""
class Split(NamedTuple):
"""The rows, their labels, and which half the loop is allowed to read."""
notes: list[str]
scores: np.ndarray
dev: np.ndarray
test: np.ndarray
# ----------------------------------------------------------------- the data
WINEMAG_CSV = (
"https://huggingface.co/datasets/GroNLP/ik-nlp-22_winemag/resolve/"
"90eb39f35fc64e556fc17f06d4137a4a69ec3297/train.csv"
)
@json_cache
def load_slice(n_dev: int, n_test: int, seed: int) -> dict:
"""Fetch the pinned CSV and take a seeded sample of note + score, one row per note."""
import csv
import io
request = urllib.request.Request(
WINEMAG_CSV, headers={"User-Agent": "typesafe-cookbook/1.0"}
)
with urllib.request.urlopen(request, timeout=300) as response:
text = response.read().decode()
rows, seen = [], set()
for row in csv.DictReader(io.StringIO(text)): # a few notes repeat verbatim
if not row["description"] or not row["points"] or row["description"] in seen:
continue
seen.add(row["description"])
rows.append((row["description"], float(row["points"])))
random.Random(seed).shuffle(rows)
picked = rows[: n_dev + n_test]
return {"notes": [r[0] for r in picked], "points": [r[1] for r in picked]}
def example_rows(split: Split, out_of_fold: np.ndarray | None, n: int) -> list[int]:
"""Select representative dev rows for a proposer round."""
dev = split.dev
if out_of_fold is None:
ranked = dev[np.argsort(split.scores[dev], kind="stable")]
return [int(ranked[round(q * (len(ranked) - 1))]) for q in np.linspace(0, 1, n)]
error = np.abs(split.scores[dev] - out_of_fold)
order = np.argsort(-error, kind="stable")
worst = [int(dev[i]) for i in order[: n // 2]]
best = [int(dev[i]) for i in order[len(order) - (n - n // 2) :]]
return worst + best
def example_block(
rows: list[int],
split: Split,
out_of_fold: np.ndarray | None,
previous: np.ndarray | None = None,
) -> str:
"""Format selected rows for the proposer."""
if out_of_fold is None:
head = "Example notes, with the score each one was given:"
body = [f"- scored {split.scores[r]:.0f}: {split.notes[r]}" for r in rows]
return head + "\n" + "\n".join(body)
head = (
"Dev notes, worst-predicted first. The first half is where your current questions "
"miss by the most and the second half is where they are already right, so what "
"separates the halves is what the questions have not captured."
)
if previous is not None:
head += (
" Each line also carries what the previous round predicted, so you can see which "
"notes your last batch of questions moved."
)
body = []
for r in rows:
line = f"- scored {split.scores[r]:.0f}, predicted {out_of_fold[r]:.1f}"
if previous is not None:
line += f" (last round {previous[r]:.1f})"
body.append(f"{line}: {split.notes[r]}")
return head + "\n" + "\n".join(body)
def load_split(n_dev: int, n_test: int, seed: int = 0) -> Split:
# keyword, because the cache key is the function name plus how each argument was spelled
data = load_slice(n_dev, n_test, seed=seed)
return Split(
notes=data["notes"],
scores=np.array(data["points"]),
dev=np.arange(n_dev),
test=np.arange(n_dev, n_dev + n_test),
)
# ----------------------------------------------------------------- step 1: propose
def proposal_prompt(examples: str, feedback: str, accepted: list[dict]) -> str:
parts = [PROPOSER_TASK, "\n" + examples]
if accepted:
parts.append(
"\nThe features you have now. `add` must not duplicate one of these; `revise` and "
"`drop` refer to one by name:\n"
+ "\n".join(
f"- {f['name']} ({f['kind']}): {f['question']}" for f in accepted
)
)
if feedback:
parts.append("\nHow the model did with those features:\n" + feedback)
return "\n".join(parts)
@json_cache
def propose(model: str, round_index: int, prompt: str) -> dict:
"""One proposal call. Every number in `prompt` is rounded so a replay hits the cache."""
if model.startswith("claude"):
import anthropic
response = anthropic.Anthropic(
api_key=os.environ.get("ANTHROPIC_API_KEY", "cache-only")
).messages.create(
model=model,
max_tokens=16000,
output_config={
"effort": "medium",
"format": {"type": "json_schema", "schema": PROPOSAL_SCHEMA},
},
messages=[{"role": "user", "content": prompt}],
)
body = next(block.text for block in response.content if block.type == "text")
usage = [response.usage.input_tokens or 0, response.usage.output_tokens or 0]
else:
from openai import OpenAI
response = OpenAI(
api_key=os.environ.get("OPENAI_API_KEY", "cache-only")
).chat.completions.create(
model=model,
reasoning_effort="high",
max_completion_tokens=16000,
response_format={"type": "json_object"},
messages=[
{
"role": "user",
"content": prompt
+ "\n\nReply with JSON matching this schema:\n"
+ json.dumps(PROPOSAL_SCHEMA),
}
],
)
body = response.choices[0].message.content
usage = [response.usage.prompt_tokens, response.usage.completion_tokens]
return {"actions": json.loads(body)["actions"][:PROPOSALS], "usage": usage}
def slug(name: str, taken: set[str]) -> str:
"""Names become question ids and column labels, so keep them plain and unique."""
base = (
"".join(c if c.isalnum() else "_" for c in name.lower()).strip("_") or "feature"
)
candidate, n = base, 2
while candidate in taken:
candidate, n = f"{base}_{n}", n + 1
return candidate
def to_candidates(actions: list[dict], accepted: list[dict], round_index: int) -> tuple:
"""Split a round's actions into screenable candidates and a list of names to drop."""
live = {f["name"] for f in accepted}
drops = [a["target"] for a in actions if a["op"] == "drop" and a["target"] in live]
replacing = {
a["target"] for a in actions if a["op"] == "revise" and a["target"] in live
}
# a revision may keep the name it replaces, since that feature is on its way out
taken, candidates = live - replacing, []
for action in actions:
if action["op"] == "drop":
continue
if action["op"] == "revise" and action["target"] not in live:
continue # a revision of something that is not there
name = slug(action["name"], taken)
taken.add(name)
candidates.append(
{
"id": f"{name}@{round_index}", # unique, so earlier rounds keep their columns
"name": name,
"kind": action["kind"],
"question": action["question"],
"replaces": action["target"] if action["op"] == "revise" else "",
}
)
return candidates, drops
# ----------------------------------------------------------------- step 2: answer
def feature_questions(features: list[dict]) -> dict:
questions = {}
for feature in features:
if feature["kind"] == "intensity":
questions[feature["name"]] = Score(
instructions=feature["question"], criteria=INTENSITY_LEVELS
)
else:
questions[feature["name"]] = Noul(
instructions=feature["question"], criteria=PRESENCE_CRITERIA
)
return questions
@json_cache
def answer(model: str, note: str, features_json: str) -> dict:
"""One request per note; every question of the round rides it. Keeps every probability."""
features = json.loads(features_json)
started = perf_counter()
response = client.system_one(
state=note, questions=feature_questions(features), model=model
)
raw = {}
for feature in features:
got = response.answers[feature["name"]]
if feature["kind"] == "intensity":
raw[feature["name"]] = [
got.probabilities.get(i, 0.0) for i in range(len(INTENSITY_LEVELS))
]
else:
raw[feature["name"]] = [got.noul]
return {
"raw": raw,
"seconds": round(perf_counter() - started, 2),
"input_tokens": response.usage.input_tokens or 0,
"output_tokens": response.usage.output_tokens or 0,
}
def featurize(notes: list[str], features: list[dict]) -> dict:
"""Answer one question set for many notes: one request each, eight in flight."""
payload = json.dumps(features, sort_keys=True)
with ThreadPoolExecutor(max_workers=8) as pool:
results = list(
pool.map(lambda note: answer(TYPESAFE_MODEL, note, payload), notes)
)
return {
f["name"]: np.array([r["raw"][f["name"]] for r in results], dtype=float)
for f in features
}
def encode(feature: dict, probabilities: np.ndarray, mode: str) -> list[tuple]:
"""Turn one question's probabilities into named columns."""
name = feature["name"]
if feature["kind"] == "presence":
return [(name, probabilities[:, 0])] # one number is all there is
levels = np.arange(probabilities.shape[1])
mean = probabilities @ levels
if mode == "mean":
return [(name, mean)]
if mode == "mean_spread":
variance = probabilities @ (levels**2) - mean**2
return [(name, mean), (f"{name}_sd", np.sqrt(np.clip(variance, 0, None)))]
return [(f"{name}_p{i}", probabilities[:, i]) for i in levels]
def design(features: list[dict], answers_for: dict, mode: str) -> tuple:
"""Stack every feature's columns into one matrix, plus a label per column."""
columns, labels = [], []
for feature in features:
for label, column in encode(feature, answers_for[feature["id"]], mode):
columns.append(column)
labels.append(label)
return np.column_stack(columns), labels
def plain(features: list[dict]) -> list[dict]:
"""What goes on the wire and into the cache key: no id, no bookkeeping."""
return [
{"name": f["name"], "kind": f["kind"], "question": f["question"]}
for f in features
]
# ----------------------------------------------------------------- step 3: fit
def rmse(y: np.ndarray, p: np.ndarray) -> float:
return float(np.sqrt(np.mean((y - p) ** 2)))
def spearman(a: np.ndarray, b: np.ndarray) -> float:
"""Rank correlation: does the model order the wines the way the critic did?"""
ranks = (
np.argsort(np.argsort(a)).astype(float),
np.argsort(np.argsort(b)).astype(float),
)
return float(np.corrcoef(*ranks)[0, 1])
def folds(y: np.ndarray, k: int, seed: int) -> list[np.ndarray]:
"""Label-stratified k-fold: sort by the label with a seeded tiebreak, then deal off the top."""
rng = np.random.default_rng(seed)
order = np.lexsort((rng.random(len(y)), y))
return [np.sort(order[i::k]) for i in range(k)]
def cross_validate(X: np.ndarray, y: np.ndarray) -> tuple[np.ndarray, float]:
out_of_fold = np.zeros((REPEATS, len(y)))
for repeat in range(REPEATS):
for fold in folds(y, FOLDS, seed=repeat):
train = np.setdiff1d(np.arange(len(y)), fold)
model = CatBoostRegressor(**CATBOOST).fit(X[train], y[train])
out_of_fold[repeat, fold] = model.predict(X[fold])
scores = [rmse(y, out_of_fold[repeat]) for repeat in range(REPEATS)]
return out_of_fold.mean(axis=0), float(np.mean(scores))
def importances(X: np.ndarray, y: np.ndarray) -> np.ndarray:
return CatBoostRegressor(**CATBOOST).fit(X, y).get_feature_importance()
def paired_gain(y: np.ndarray, before: np.ndarray, after: np.ndarray) -> tuple:
"""Bootstrap the paired held-out RMSE change."""
squared = ((y - before) ** 2, (y - after) ** 2)
rng = np.random.default_rng(0)
drawn = []
for _ in range(2000):
rows = rng.integers(0, len(y), len(y))
drawn.append(
np.sqrt(squared[1][rows].mean()) - np.sqrt(squared[0][rows].mean())
)
drawn = np.array(drawn)
return (
rmse(y, after) - rmse(y, before),
float(np.percentile(drawn, 2.5)),
float(np.percentile(drawn, 97.5)),
)
def fit_predict(X: np.ndarray, split: Split) -> np.ndarray:
model = CatBoostRegressor(**CATBOOST).fit(X[split.dev], split.scores[split.dev])
return model.predict(X[split.test])
def fit_predict_text(split: Split) -> np.ndarray:
"""The reference arm: the same model, handed the note instead of the columns."""
from catboost import Pool
raw = np.array([[note] for note in split.notes], dtype=object)
model = CatBoostRegressor(**CATBOOST).fit(
Pool(raw[split.dev], split.scores[split.dev], text_features=[0])
)
return model.predict(Pool(raw[split.test], text_features=[0]))
def evaluate(
features: list[dict], answers_for: dict, split: Split, mode: str
) -> tuple[np.ndarray, float]:
"""Cross-validated error on the dev rows for one candidate question set."""
X, _ = design(features, answers_for, mode)
return cross_validate(X[split.dev], split.scores[split.dev])
def swap_in(accepted: list[dict], feature: dict) -> list[dict] | None:
"""The accepted set with `feature` in place of the one it revises, or None if it is gone."""
at = next(
(i for i, f in enumerate(accepted) if f["name"] == feature["replaces"]), None
)
if at is None:
return None
trial = list(accepted)
trial[at] = {k: feature[k] for k in ("id", "name", "kind", "question")}
return trial
def try_change(
trial: list[dict],
accepted: list[dict],
cv: float,
answers_for: dict,
split: Split,
mode: str,
tolerance: float,
) -> tuple[list[dict], float, str, bool]:
"""Refit with the change and keep it only if the dev error improves. No API calls."""
_, cv_trial = evaluate(trial, answers_for, split, mode)
if cv_trial <= cv + tolerance:
return trial, cv_trial, f"CV {cv:.3f} -> {cv_trial:.3f}", True
return accepted, cv, f"would cost {cv_trial - cv:+.3f}", False
def owner_of(label: str, features: list[dict]) -> dict:
"""Which feature a column label belongs to - encodings suffix the name."""
exact = next((f for f in features if f["name"] == label), None)
if exact:
return exact
return next(f for f in features if label.startswith(f["name"] + "_"))
def importance_per_feature(
features: list[dict], labels: list[str], column_importances: np.ndarray
) -> dict:
"""Sum each question's CatBoost column importances.
Intensity questions can produce multiple model columns. Combining their normalized
importances gives one percentage share per question.
"""
total = {f["name"]: 0.0 for f in features}
for label, column_importance in zip(labels, column_importances):
total[owner_of(label, features)["name"]] += float(column_importance)
return total
def feedback_for(
history: list[float],
accepted: list[dict],
answers_for: dict,
split: Split,
mode: str,
out_of_fold: np.ndarray,
previous: np.ndarray | None,
) -> str:
"""The scoreboard the next proposal call reads. The notes themselves arrive separately,
through `example_block`. Numbers are rounded before they enter the prompt."""
X, labels = design(accepted, answers_for, mode)
dev, scores = split.dev, split.scores
by_name = importance_per_feature(accepted, labels, importances(X[dev], scores[dev]))
lines = ["Cross-validated RMSE in points so far, lower is better:"]
lines += [f" round {i + 1}: {v:.2f}" for i, v in enumerate(history)]
if previous is not None:
now, before = np.abs(scores[dev] - out_of_fold), np.abs(scores[dev] - previous)
better, worse = int((now < before - 0.1).sum()), int((now > before + 0.1).sum())
lines.append(
f"\nAgainst the previous round, {better} of the {len(dev)} dev notes are now "
f"predicted better by more than 0.1 points and {worse} are predicted worse."
)
lines.append(
"\nYour features, with importance as a percentage of the total and the spread of the "
"column across the dev rows. Low importance or low spread means the question is not "
"doing much; revise or drop it."
)
for feature in sorted(accepted, key=lambda f: -by_name.get(f["name"], 0.0)):
column = encode(feature, answers_for[feature["id"]], mode)[0][1]
lines.append(
f" {feature['name']} ({feature['kind']}): "
f"{by_name.get(feature['name'], 0.0):.1f}% importance, "
f"spread {column[dev].std():.2f}"
)
return "\n".join(lines)
# ----------------------------------------------------------------- the loop itself
class Discovery(NamedTuple):
"""Artifacts returned by the discovery loop."""
accepted: list[dict] # the question set it ended with
answers_for: dict # feature id -> (rows x levels) probabilities
snapshots: list[list[dict]] # the set as it stood at the end of each round
history: list[float] # dev CV error after each round
batches: list[tuple] # what each round sent, for the request table
journal: list[tuple] # every action and what became of it
def run_loop(
split: Split,
proposer: str,
rounds: int,
examples: int,
mode: str,
min_spread: float,
tolerance: float,
) -> Discovery:
"""Run the propose, answer, fit, and feedback loop."""
shown = example_rows(split, None, examples) # round 1 has nothing predicted yet
out_of_fold = previous = None
got_from = Discovery([], {}, [], [], [], [])
accepted, answers_for = got_from.accepted, got_from.answers_for
snapshots, history = got_from.snapshots, got_from.history
batches, journal = got_from.batches, got_from.journal
feedback = ""
for round_index in range(1, rounds + 1):
block = example_block(shown, split, out_of_fold, previous)
actions = propose(
proposer, round_index, proposal_prompt(block, feedback, accepted)
)["actions"]
keep, drops = to_candidates(actions, accepted, round_index)
if keep: # one request per row, carrying every question this round proposed
batches.append((round_index, plain(keep)))
answers = featurize(split.notes, plain(keep))
for feature in keep:
answers_for[feature["id"]] = answers[feature["name"]]
for (
feature
) in keep: # an add goes in; importance says later whether it earned it
if feature["replaces"]:
continue
column = encode(feature, answers_for[feature["id"]], mode)[0][1]
flat = float(column[split.dev].std()) < min_spread
journal.append(
(round_index, "flat" if flat else "add", feature["name"], "")
)
if not flat:
accepted.append(
{k: feature[k] for k in ("id", "name", "kind", "question")}
)
_, cv = evaluate(accepted, answers_for, split, mode)
trial_args = (answers_for, split, mode, tolerance)
for feature in [f for f in keep if f["replaces"]]: # every revision is tried
trial = swap_in(accepted, feature)
if trial is None: # it revises something an earlier round already dropped
journal.append(
(round_index, "stale", feature["name"], "target is gone")
)
continue
accepted[:], cv, note, took = try_change(trial, accepted, cv, *trial_args)
what = "revise" if took else "reject"
journal.append(
(
round_index,
what,
feature["name"],
f"was {feature['replaces']}, {note}",
)
)
for name in drops: # and so is every drop
trial = [f for f in accepted if f["name"] != name]
if not trial:
continue
accepted[:], cv, note, took = try_change(trial, accepted, cv, *trial_args)
journal.append((round_index, "drop" if took else "keep", name, note))
previous, (out_of_fold, cv) = (
out_of_fold,
evaluate(accepted, answers_for, split, mode),
)
history.append(cv)
snapshots.append(list(accepted))
feedback = feedback_for(
history, accepted, answers_for, split, mode, out_of_fold, previous
)
# next round reads the rows these questions get most wrong, and as many they get right
shown = example_rows(split, out_of_fold, examples)
report(round_index, keep, drops, journal, accepted, cv)
return got_from
def report(
round_index: int,
keep: list[dict],
drops: list[str],
journal: list[tuple],
accepted: list[dict],
cv: float,
) -> None:
"""One block per round: the counts, the names it added, then everything with a number."""
revised = sum(1 for f in keep if f["replaces"])
print(
f"round {round_index}: {len(keep) - revised} add, {revised} revise, "
f"{len(drops)} drop"
)
this_round = [j for j in journal if j[0] == round_index]
added = [name for _, what, name, _ in this_round if what == "add"]
if added:
print(
textwrap.fill(
", ".join(added),
88,
initial_indent=" added ",
subsequent_indent=" " * 10,
)
)
for _, what, name, note in this_round: # everything carrying a number of its own
if what != "add":
print(f" {what:<7}{name:<34}{note}")
print(f" -> {len(accepted)} features, dev CV RMSE {cv:.3f}\n")
# ----------------------------------------------------------------- asking for the score
@json_cache
def ask_score(model: str, note: str) -> dict:
"""One `Score` over ten quality bands, read as a level and rescaled to 80-100."""
response = client.system_one(
state=note,
questions={
"quality": Score(
instructions=(
"Judging only by what this tasting note says, how good is the wine?"
),
criteria=SCORE_LEVELS,
)
},
model=model,
)
got = response.answers["quality"]
top = len(SCORE_LEVELS) - 1
expected = sum(k * v for k, v in got.probabilities.items())
return {
# level 0 is the bottom of the critic's scale, level 9 the top
"expected": 80.0 + 20.0 * expected / top,
"picked": 80.0 + 20.0 * got.score / top,
"input_tokens": response.usage.input_tokens or 0,
"output_tokens": response.usage.output_tokens or 0,
}
# ----------------------------------------------------------------- charts
SURFACE, INK, INK2, MUTED = "#fcfcfb", "#0b0b0b", "#52514e", "#898781"
GRID, AXIS, BLUE, ORANGE = "#e1e0d9", "#c3c2b7", "#2a78d6", "#eb6834"
def style(ax) -> None:
ax.set_facecolor(SURFACE)
for side in ("top", "right"):
ax.spines[side].set_visible(False)
for side in ("left", "bottom"):
ax.spines[side].set_color(AXIS)
ax.tick_params(colors=MUTED, labelcolor=INK2, labelsize=9)
ax.set_axisbelow(True)
def polarity(feature: dict, answers_for: dict, split: Split) -> float:
"""Rank correlation between a question's answer and the critic score, on the dev rows.
Positive means a higher answer goes with a better review, negative the opposite. It is
what orders the rows of the feature map, so the map reads as a gradient that flips.
"""
column = encode(feature, answers_for[feature["id"]], "mean")[0][1]
return spearman(column[split.dev], split.scores[split.dev])
def reviews_heatmap(
plt,
questions: list[dict],
answers_for: dict,
split: Split,
rows: tuple,
):
"""Compare held-out reviews across the discovered questions, best-signal first.
Rows arrive sorted from the questions that rise with the score to the ones that fall with
it, so a row above the divider shades left to right and a row below it shades right to
left.
"""
def value_of(feature: dict, row: int) -> float:
return float(encode(feature, answers_for[feature["id"]], "mean")[0][1][row])
signs = [polarity(question, answers_for, split) for question in questions]
flip = next((i for i, s in enumerate(signs) if s < 0), len(questions))
raw = np.array(
[[value_of(question, row) for row in rows] for question in questions]
)
normalized = np.array(
[
values / (4 if question["kind"] == "intensity" else 1)
for question, values in zip(questions, raw)
]
)
cmap = matplotlib.colors.LinearSegmentedColormap.from_list(
"typesafe_heat", [SURFACE, "#f7c7ad", ORANGE]
)
fig, ax = plt.subplots(
figsize=(9.5, 1.8 + 0.58 * len(questions)), facecolor=SURFACE
)
image = ax.imshow(normalized, aspect="auto", cmap=cmap, vmin=0, vmax=1)
row_labels = []
for question, sign in zip(questions, signs):
kind = "score" if question["kind"] == "intensity" else "noul"
prefix = f"{sign:+.2f} ({kind}) "
lines = textwrap.wrap(
" ".join(question["question"].split()),
width=52,
max_lines=2,
placeholder="...",
break_long_words=False,
break_on_hyphens=False,
)
row_labels.append(prefix + (f"\n{' ' * len(prefix)}").join(lines))
column_labels = [
f"#{i}\n{split.scores[row]:.0f} points\n{' '.join(split.notes[row].split())[:15]}..."
for i, row in enumerate(rows, 1)
]
ax.set_yticks(np.arange(len(questions)), row_labels)
ax.set_xticks(np.arange(len(rows)), column_labels)
ax.tick_params(
axis="x", top=True, labeltop=True, bottom=False, labelbottom=False, pad=8
)
ax.tick_params(axis="y", labelsize=8.5)
for side in ax.spines.values():
side.set_visible(False)
ax.set_xticks(np.arange(-0.5, len(rows), 1), minor=True)
ax.set_yticks(np.arange(-0.5, len(questions), 1), minor=True)
ax.grid(which="minor", color=SURFACE, linewidth=2)
ax.tick_params(which="minor", bottom=False, left=False)
for i, question in enumerate(questions):
for j, value in enumerate(raw[i]):
label = (
f"{value:.1f}" if question["kind"] == "intensity" else f"{value:.2f}"
)
color = SURFACE if normalized[i, j] > 0.58 else INK2
ax.text(j, i, label, ha="center", va="center", color=color, fontsize=8)
# the line where the questions stop rising with the score and start falling with it
if 0 < flip < len(questions):
ax.axhline(flip - 0.5, color=INK, linewidth=1.2)
ax.annotate(
"a higher answer means a worse review, below this line",
(len(rows) - 0.5, flip - 0.5),
xytext=(-4, 5),
textcoords="offset points",
va="bottom",
ha="right",
color=INK2,
fontsize=8.5,
)
colorbar = fig.colorbar(image, ax=ax, fraction=0.025, pad=0.025)
colorbar.set_ticks([0, 0.5, 1])
colorbar.set_label("normalized answer", color=INK2, fontsize=8.5)
colorbar.ax.tick_params(labelsize=8, colors=INK2)
fig.suptitle(
"Every question, on five held-out reviews from worst to best",
x=0.01,
y=0.995,
ha="left",
color=INK,
fontsize=11,
)
fig.text(
0.01,
0.972,
"sorted by how the answer moves with the score, so each row above the line shades "
"left to right and each row below it shades the other way",
color=MUTED,
fontsize=9,
)
fig.text(
0.01,
0.005,
"Row labels lead with the rank correlation between that question's answer and the "
"critic score. Cell text is each question's native scale: score 0-4, noul 0-1.",
color=MUTED,
fontsize=8.5,
)
return fig
def rounds_chart(
plt, curve: list[tuple], history: list[float], n_test: int, gain: tuple
):
"""Dev error and held-out error per round. The trend is the point, not the gap."""
rounds = list(range(1, len(curve) + 1))
values = [v for _, v in curve]
fig, ax = plt.subplots(figsize=(7, 3.9), facecolor=SURFACE)
style(ax)
ax.grid(axis="y", color=GRID, linewidth=0.8)
# each dev fold trains on four fifths of the rows, so the dev line sits the higher of the two
ax.fill_between(rounds, history, values, color=GRID, alpha=0.75, linewidth=0)
ax.plot(
rounds,
history,
marker="o",
color=BLUE,
linewidth=2,
linestyle="--",
label="dev, cross-validated - what the loop optimises",
)
ax.plot(
rounds,
values,
marker="o",
color=ORANGE,
linewidth=2,
label="held out - what that actually buys",
)
# label each point on the outside of the pair, so neither line crowds its own numbers
for x, dev_value, test_value in zip(rounds, history, values):
for value, other in ((dev_value, test_value), (test_value, dev_value)):
ax.annotate(
f"{value:.2f}",
(x, value),
textcoords="offset points",
xytext=(0, 8 if value >= other else -16),
ha="center",
color=INK2,
fontsize=8.5,
)
ax.set_xticks(
rounds, [f"round {x}\n{n} features" for x, (n, _) in zip(rounds, curve)]
)
ax.set_ylabel("RMSE in points (lower is better)", color=INK2, fontsize=9)
# tight around the two lines: the whole finding lives inside 0.15 of a point
low, high = min(values + history), max(values + history)
ax.set_ylim(low - 0.10, high + 0.05)
difference, low_ci, high_ci = gain
ax.set_title(
f"{len(rounds)} rounds of the loop, scored on {n_test} held-out reviews",
loc="left",
color=INK,
fontsize=11,
pad=20,
)
# the number the chart is really about: is the held-out move bigger than the noise?
ax.text(
0,
1.015,
f"round 1 to round {len(rounds)}, held out: {difference:+.3f} points, "
f"95% CI [{low_ci:+.3f}, {high_ci:+.3f}]",
transform=ax.transAxes,
color=MUTED,
fontsize=9,
)
ax.legend(frameon=False, labelcolor=INK2, fontsize=9, loc="lower left")
return figInstallation
pip install anthropic openai catboost numpy matplotlib ipython "typesafe-sdk>=0.5.7" cooksafe --extra-index-url https://pypi.typesafe.ai/puis définis TYPESAFE_API_KEY et ANTHROPIC_API_KEY. Chaque appel API est mis en cache dans
json_cache.json, qui est livré avec le cookbook, donc un nouveau rendu rejoue ces nombres
sans rien appeler. Supprime-le pour relancer en direct. Les nombres proviennent de TypeSafe
jev-1.12 et claude-sonnet-5 le 2026-08-03. propose() possède une deuxième branche pour
gpt-5.6-luna, qui n’a pas été exécutée.
La première cellule de code est l’implémentation complète : appels API, encodages, métriques, style des graphiques. Elle est là pour que ce fichier s’exécute seul, et le site de documentation la masque. Passez-la sous silence lors d’une première lecture — la recette commence juste en dessous.
N_DEV, N_TEST = 1200, 800 # the loop reads dev labels only; test is scored once
ROUNDS = 5 # a round answers questions for all 2,000 rows: 2,000 requests
PROPOSER = "claude-sonnet-5" # or "gpt-5.6-luna"; the cache holds the Anthropic run
EXAMPLES = 60 # dev notes the proposer reads per round, half of them its worst misses
MIN_SPREAD = 0.05 # a column this flat cannot separate anything, so it is not kept
CHANGE_TOLERANCE = 0.0 # a revision or drop has to improve dev error, not just not hurt
ENCODING = "mean_spread" # a score answer becomes two columns: its mean and spread
split = load_split(N_DEV, N_TEST, seed=0)
NOTES, SCORES, DEV, TEST = split.notes, split.scores, split.dev, split.test
print(
f"{len(DEV)} dev rows, {len(TEST)} held out; scores run "
f"{SCORES.min():.0f}-{SCORES.max():.0f}, mean {SCORES.mean():.2f}, sd {SCORES.std():.2f}"
)
print(f"\none of the notes:\n{NOTES[0]}")1200 dev rows, 800 held out; scores run 80-98, mean 88.73, sd 3.17
one of the notes:
A Champagne that is very much wine. The structure and the richness are just right for a food wine, showing ripe acidity, flavors of plums and apricots, and balancing these primary fruits with a dense, complex structure that takes in yeast, maturity and a tight apple skin finish.La boucle lit les mêmes 1 200 lignes sur les 2 000 (les lignes de développement) encore et encore, et conserve une question lorsqu’elle aide à prédire ces 1 200 scores. Évaluer sur les mêmes lignes mesurerait surtout la capacité de la boucle à s’ajuster à elles, aussi les 800 autres sont conservées à part et évaluées une seule fois, à la fin.
Deux types de questions
Une question proposée est de l’une des deux sortes, et la sorte détermine quel nombre revient.
intensitydevient unScore, pour tout ce qui s’exprime en degrés. Ses cinq niveaux sont imprimés ci-dessous, et la colonne correspond au niveau moyen, de sorte qu’une note située entre « modéré » et « fortement » s’affiche entre les deux.presencedevient unNoul, pour un fait binaire (oui/non) tel que la présence ou l’absence d’un nom de défaut. La colonne indique cette probabilité.
La méthode
questions <- {}
repeat for each round:
notes <- round 1 ? 60 dev notes across the score range
: the 30 worst-predicted dev notes + the 30 best,
each with its score, this prediction and the last
actions <- LLM(brief, questions, notes, importance and error so far)
answers[q] <- TypeSafe(note, all new questions of this round) for every row
for each added q: keep it unless its column is flat
for each revised q: refit; keep the change only if dev error drops
for each dropped q: refit; drop it only if dev error drops
out_of_fold <- k-fold CatBoost on the columns # judges, and picks next round's notesAucune question n’est filtrée avant d’être répondue. Toutes les questions d’un tour sont envoyées dans la même requête, donc ajouter une question ne coûte pas de requête supplémentaire. Une question qui ne s’applique qu’à une ligne sur dix paraîtra inutile dans les 60 notes que le proposant lit, tout en restant la colonne la plus utile de l’ensemble.
k-fold signifie diviser les lignes de dev en k parties et prédire chaque partie avec un modèle entraîné sur les autres parties. Ces prédictions remplissent trois fonctions : elles jugent chaque révision et suppression, elles sélectionnent les notes que le tour suivant lit, et elles indiquent au proposant lesquelles de ses questions ont aidé, de combien elles ont évolué depuis le tour précédent.
print("every intensity question is graded on these five levels:\n")
for i, level in enumerate(INTENSITY_LEVELS):
print(f" {i}. {level}")
print("\nevery presence question is judged true or false against these:\n")
print(f" true: {PRESENCE_CRITERIA['true']}")
print(f" false: {PRESENCE_CRITERIA['false']}")
print("\nthe brief the proposer works from:\n")
print("\n".join(PROPOSER_TASK.splitlines()[:6]) + "\n ...")every intensity question is graded on these five levels:
0. Not present in this note at all
1. Barely present - mentioned once, in passing
2. Present at a moderate level
3. Present strongly - the note dwells on it
4. Dominant - the note is largely about this
every presence question is judged true or false against these:
true: The note states this or clearly implies it
false: The note gives no indication of this
the brief the proposer works from:
You are designing numeric features for a gradient-boosting model that
predicts the score a wine critic gave (an integer from 80 to 100) from the tasting note alone.
The model sees nothing but the features you design.
Return up to 18 actions. Each action is one of:
...La boucle d’auto-recherche
run_loop exécute les cinq tours et affiche un bloc par tour. Une question ajoutée est insérée directement : ses réponses ont déjà été récupérées, et son importance montrera plus tard si elle valait la peine d’être posée. Une révision ou un retrait enlève une colonne que le modèle utilise déjà, donc chacun est essayé en premier : recalibrage avec le changement, et conservation uniquement si l’erreur de validation diminue. Un recalibrage ne coûte aucun appel API, donc essayer un changement et le rejeter est gratuit.
run = run_loop(
split, PROPOSER, ROUNDS, EXAMPLES, ENCODING, MIN_SPREAD, CHANGE_TOLERANCE
)
accepted, answers_for = run.accepted, run.answers_for
snapshots, history = run.snapshots, run.historyround 1: 18 add, 0 revise, 0 drop
added complexity, fruit_intensity, tannin_structure, acidity_intensity,
oak_intensity, finish_length, balance_harmony, aging_potential,
positive_superlative_language, negative_critical_language,
drinkability_easiness, body_richness, sweetness_level, texture_descriptors,
earthy_savory_notes, flaw_or_defect_mentioned,
single_vineyard_or_prestige_signal, varietal_blend_detail
-> 18 features, dev CV RMSE 1.903
round 2: 5 add, 3 revise, 3 drop
added power_concentration_language, flavor_distinctiveness, generic_fruit_language,
candied_artificial_flavor, rustic_authentic_character
reject oak_dominance was oak_intensity, would cost +0.005
revise negative_critical_language was negative_critical_language, CV 1.897 -> 1.894
revise single_vineyard_or_prestige_signalwas single_vineyard_or_prestige_signal, CV 1.894 -> 1.881
keep finish_length would cost +0.009
keep texture_descriptors would cost +0.001
keep varietal_blend_detail would cost +0.023
-> 23 features, dev CV RMSE 1.881
round 3: 7 add, 2 revise, 1 drop
added elegance_finesse_language, minerality_precision_language,
hedged_qualified_praise, underripe_green_character,
reviewer_overall_verdict_strength, unusual_or_funky_descriptor_valence,
botrytis_or_special_winemaking_signal
revise negative_critical_language was negative_critical_language, CV 1.868 -> 1.864
revise finish_quality was finish_length, CV 1.864 -> 1.861
keep candied_artificial_flavor would cost +0.014
-> 30 features, dev CV RMSE 1.861
round 4: 5 add, 2 revise, 3 drop
added excess_or_imbalance_signal, descriptive_detail_density,
critic_enthusiasm_confidence, savory_food_wine_seriousness,
note_overall_tone_positivity
revise rustic_authentic_character was rustic_authentic_character, CV 1.843 -> 1.838
reject hedged_qualified_praise was hedged_qualified_praise, would cost +0.014
keep botrytis_or_special_winemaking_signalwould cost +0.011
keep candied_artificial_flavor would cost +0.009
keep unusual_or_funky_descriptor_valencewould cost +0.010
-> 35 features, dev CV RMSE 1.838
round 5: 4 add, 2 revise, 8 drop
added structural_seriousness, youthful_tension_signal, surface_prettiness_vs_depth,
price_value_signal
reject unconventional_character_as_virtuewas rustic_authentic_character, would cost +0.010
revise flavor_distinctiveness was flavor_distinctiveness, CV 1.849 -> 1.843
keep candied_artificial_flavor would cost +0.002
keep botrytis_or_special_winemaking_signalwould cost +0.003
keep hedged_qualified_praise would cost +0.006
keep excess_or_imbalance_signal would cost +0.005
drop underripe_green_character CV 1.843 -> 1.840
keep unusual_or_funky_descriptor_valencewould cost +0.002
keep texture_descriptors would cost +0.001
keep generic_fruit_language would cost +0.000
-> 38 features, dev CV RMSE 1.840Pointer vers vos propres données
PROPOSER_TASK est la seule chaîne qui mentionne le vin, et featurize() prend n’importe quelle liste de chaînes. Modifier ce bref change le prompt de proposition, et le prompt fait partie de la clé de cache, donc le prochain run appelle l’API à nouveau pour chaque tour.
Le nombre de requêtes augmente avec le nombre de lignes, pas avec le nombre de questions : une requête par ligne par tour, donc 100 000 lignes correspondent à 100 000 requêtes par tour. Une révision compte comme une nouvelle question, ce qui implique un autre passage sur toutes les lignes. Augmentez progressivement le pool de travailleurs. Huit est déjà suffisant pour atteindre une limite de taux sur une clé partagée.
Ce que les questions voient
Cinq évaluations hors échantillon, une à chaque quart de la plage de score, confrontées à quinze des 38 questions : les huit questions de score les plus importantes par ordre d’importance, plus les sept nouls les plus importants.
Ces quinze lignes sont ensuite triées selon la direction dans laquelle la réponse évolue avec le score du critique. Les questions dont la réponse augmente avec le score arrivent en premier, celles dont la réponse diminue avec celui-ci se placent après le séparateur. Ainsi, en allant de gauche à droite, de la critique la plus mauvaise à la meilleure, les réponses situées au-dessus du séparateur devraient croître et celles situées en dessous devraient décroître.
X, labels = design(accepted, answers_for, ENCODING)
column_importances = importances(X[DEV], SCORES[DEV])
# an encoding gives a feature more than one column, so add a feature's columns back up
feature_importances = importance_per_feature(accepted, labels, column_importances)
ranked = sorted(accepted, key=lambda f: -feature_importances[f["name"]])
score_questions = [f for f in ranked if f["kind"] == "intensity"][:8]
noul_questions = [f for f in ranked if f["kind"] == "presence"][:7]
# ordered by which way the answer moves with the score, so the map flips halfway down
heatmap_questions = sorted(
score_questions + noul_questions,
key=lambda f: -polarity(f, answers_for, split),
)
ordered_test = TEST[np.argsort(SCORES[TEST], kind="stable")]
positions = np.linspace(0, len(ordered_test) - 1, 5).round().astype(int)
review_rows = tuple(ordered_test[positions])
print("the five held-out heatmap columns:\n")
for i, row in enumerate(review_rows, 1):
excerpt = " ".join(NOTES[row].split())
print(f" {i}. {SCORES[row]:.0f} points: {excerpt[:100]}...")
fig = reviews_heatmap(plt, heatmap_questions, answers_for, split, review_rows)
display(fig)
plt.close(fig)the five held-out heatmap columns:
1. 80 points: Raw cherry and plum aromas are resiny and suggest wet cement. This is shearing and so jacked up with...
2. 86 points: A slight spritz brightens the mouthfeel of this lemony wine. Aromas are a bit musky, but flavors of ...
3. 89 points: This is a European-style Syrah, cofermented with 2% Viognier. It's soft and round, medium in body, a...
4. 91 points: From the producer's dry-farmed estate vineyard, and supported by small amounts of Merlot and Caberne...
5. 97 points: A thoroughly elegant, serious and yet immensely enjoyable wine that stays lively many days after ope...
Le tableau du haut de la page, calculé. Les cinq bras sont évalués une fois sur les mêmes 800
lignes conservées, et les trois premiers sautent la découverte de fonctionnalités. L’un prédit la moyenne des scores de dev
et ne lit rien de la note du tout. L’autre transmet la note au même CatBoost
via sa gestion de text_features, ce qui la transforme en compte de mots. L’un demande à TypeSafe
le score lui-même.
Ce troisième élément correspond à un Score unique par ligne sur dix bandes de qualité, allant de « défectueux ou désagréable » à « profond ». Dix, car dix niveaux constituent le maximum qu’une question Score peut prendre ; au-delà, le onzième renvoie une erreur serveur. Le niveau 0 correspond à 80 points et le niveau 9 à 100. Répartir les bandes sur l’échelle de cette manière ne suffit pas en soi, car rien dans la question n’indique où se situent réellement les scores de cette publication sur cette échelle. Ainsi, chaque réponse est ensuite décalée d’un seul offset, mesuré sur les scores de développement. Cet offset est imprimé dans l’étiquette de ligne, et c’est la seule chose que ce raccourci apprend des scores.
Spearman est la corrélation de rang, où 1,0 placerait les vins hors test exactement dans l’ordre du critique. La ligne de nombre de mots correspond au traitement de texte propre à CatBoost, et non à un pipeline de régression de texte optimisé. Tout cela constitue un seul jeu de données et une seule exécution de la boucle.
predicted = fit_predict(X, split)
text_predicted = fit_predict_text(split)
# ask TypeSafe for the score itself, one request per row
with ThreadPoolExecutor(max_workers=8) as pool:
direct = list(pool.map(lambda note: ask_score(TYPESAFE_MODEL, note), NOTES))
asked = np.array([d["expected"] for d in direct])
shift = float(SCORES[DEV].mean() - asked[DEV].mean()) # one number, from the dev labels
# what one proposal call gets you, before any feedback: the set round 1 ended with
first_round, _ = design(snapshots[0], answers_for, ENCODING)
print(f"{'arm':<46}{'RMSE':>7}{'spearman':>10}")
for label, p in (
("predict the mean of the dev rows", np.full(len(TEST), SCORES[DEV].mean())),
("the note as word counts, same CatBoost", text_predicted),
(f"ask for the score itself, shifted {shift:+.2f}", asked[TEST] + shift),
(
f"{len(snapshots[0])} questions from round 1, no loop",
fit_predict(first_round, split),
),
(f"{len(accepted)} questions after all {ROUNDS} rounds", predicted),
):
print(f"{label:<46}{rmse(SCORES[TEST], p):>7.3f}{spearman(SCORES[TEST], p):>10.3f}")arm RMSE spearman
predict the mean of the dev rows 3.088 -0.014
the note as word counts, same CatBoost 2.466 0.605
ask for the score itself, shifted -1.71 2.145 0.761
18 questions from round 1, no loop 1.869 0.778
38 questions after all 5 rounds 1.772 0.799Les tours d’autorecherche ont-ils été utiles ?
Les deux courbes représentent l’erreur de l’ensemble de questions à la fin de chaque tour, en partant de la première proposition. La ligne en pointillés indique l’erreur de validation croisée sur les données de développement, sur laquelle chaque décision d’acceptation ou de rejet est fondée. La ligne pleine évalue le même ensemble de questions sur les lignes conservées hors boucle, que la boucle ne consulte jamais. Chaque point correspond à l’état de l’ensemble au moment de la clôture du tour ; ainsi, un tour qui n’a fait que réviser ou supprimer une question fait tout de même avancer les deux courbes. La carte des caractéristiques précise ce que mesurent les questions ; l’erreur est ce qui permet de savoir si les tours suivant la première proposition ont amélioré les prédictions.
L’axe est serré : tout ce qui s’y produit se fait en un cinquième de point, et chaque raccourci du tableau ci-dessus se situe bien au-dessus de celui-ci. La ligne dev se trouve au-dessus de la ligne tenue à l’écart tout au long, et cela est un effet de taille d’entraînement. Chaque pli dev s’entraîne sur quatre cinquièmes des lignes dev, tandis que le nombre tenu à l’écart provient d’un modèle qui a reçu les 1 200. Les deux lignes évoluent ensemble, de sorte que le nombre dev par lequel la boucle oriente le suivi correspond au nombre tenu à l’écart qu’il ne voit jamais. L’intervalle sous le titre provient du rééchantillonnage des lignes tenues à l’écart, il indique donc si le passage du tour 1 au tour 5 est plus important que le bruit dans 800 lignes.
curve, per_round = [], []
for features in snapshots:
X_round, _ = design(features, answers_for, ENCODING)
per_round.append(fit_predict(X_round, split))
curve.append((len(features), rmse(SCORES[TEST], per_round[-1])))
# the same held-out rows resampled 2,000 times, both arms scored on each resample
gain = paired_gain(SCORES[TEST], per_round[0], per_round[-1])
print(
f"round 1 -> round {ROUNDS} on the held-out rows: {gain[0]:+.3f} points, "
f"95% CI [{gain[1]:+.3f}, {gain[2]:+.3f}]"
)
fig = rounds_chart(plt, curve, history, len(TEST), gain)
display(fig)
plt.close(fig)round 1 -> round 5 on the held-out rows: -0.097 points, 95% CI [-0.147, -0.050]
La ligne hors-échantillonnage chute davantage que la ligne de développement. Le tour 1 a rédigé ses questions sans retour d’information pour s’appuyer dessus, et les quatre tours suivants valent 0,10 point sur les lignes hors-échantillonnage, IC à 95 % [-0,147, -0,050].
Le round 5 a proposé quatre ajouts, deux reformulations et huit suppressions, et a donné le premier numéro de développement qui ne s’est pas amélioré. On ne peut pas demander autant qu’à propos d’une note de 245 caractères, et au round 5, les propositions ont basculé de l’ajout de questions à leur suppression.
kinds = {f["name"]: f["kind"] for f in accepted}
print("feature importance share: % of total CatBoost importance across all questions")
print(f"{'feature':<38}{'asked as':<10}{'importance share':>16}")
for name, importance_share in sorted(feature_importances.items(), key=lambda p: -p[1])[
:12
]:
kind = "score" if kinds[name] == "intensity" else "noul"
print(
f"{name[:36]:<38}{kind:<10}{importance_share:>8.1f}% "
f"{'#' * round(importance_share)}"
)
counts = f"{sum(1 for k in kinds.values() if k == 'intensity')} score"
counts += f", {sum(1 for k in kinds.values() if k == 'presence')} noul"
print(f"\nthe {len(accepted)} questions the loop kept: {counts}")
top = max(feature_importances, key=feature_importances.get)
print(
f'the question behind the top row:\n {top}: "{owner_of(top, accepted)["question"]}"'
)feature importance share: % of total CatBoost importance across all questions
feature asked as importance share
note_overall_tone_positivity score 17.4% #################
savory_food_wine_seriousness score 8.7% #########
positive_superlative_language score 8.4% ########
single_vineyard_or_prestige_signal noul 7.2% #######
descriptive_detail_density score 5.7% ######
elegance_finesse_language score 5.0% #####
complexity score 5.0% #####
aging_potential score 5.0% #####
balance_harmony score 2.9% ###
drinkability_easiness score 2.9% ###
critic_enthusiasm_confidence score 2.7% ###
flavor_distinctiveness score 2.6% ###
the 38 questions the loop kept: 29 score, 9 noul
the question behind the top row:
note_overall_tone_positivity: "Setting aside specific descriptors, how positive is the overall emotional tone and word choice of the note taken as a whole (warm, admiring language throughout vs. flat, neutral, or lukewarm phrasing)?"importance share représente l’importance des caractéristiques CatBoost, normalisée de sorte que les 38 questions totalisent 100 %. Il ne s’agit pas d’une part de lignes, de questions ou de précision de prédiction. Une question de score possède deux colonnes, une moyenne et une dispersion, de sorte que les importances de ses deux colonnes sont additionnées avant l’affichage du pourcentage. note_overall_tone_positivity représente 17,4 % du total. La quatrième ligne est un noul : le fait que la note nomme un seul vignoble ou un autre signal de prestige est une information oui/non, elle a donc été posée sous forme de question unique.
Prochaines étapes
Cette exécution garde la boucle petite. Extensions directes :
- Évaluer un candidat avant de payer pour sa réponse. Traitez la question proposée elle-même comme l’état et interrogez les nouls à son sujet : peut-elle être répondue à partir du texte source, signifie-t-elle une seule chose selon ses critères, s’applique-t-elle à la plupart des lignes, variera-t-elle d’une ligne à l’autre. N’envoyez que les questions qui passent les quatre critères avec une confiance suffisante.
- Élaguer les fonctionnalités corrélées. Mesurez la corrélation entre les colonnes encodées sur les lignes de développement, regroupez les quasi-doublons, et conservez la question la plus claire ou la plus importante de chaque groupe.
- Ajouter des références simples. Comparez TF-IDF, les comptes de caractères et d’autres fonctionnalités structurelles prises individuellement, puis ajoutez-les aux colonnes découvertes pour mesurer la contribution de chacune.
- Mélanger les familles de proposeurs. Générez des lots de candidats avec Anthropic, OpenAI, Google Gemini et des modèles open source, puis fusionnez-les et supprimez les doublons avant qu’ils n’atteignent TypeSafe. Différentes familles devraient élargir la recherche davantage que des appels répétés à un seul proposeur.
- Comparer les modèles et méthodes prédictifs. Essayez la régression linéaire ou élastique-net, un régresseur à vecteurs de support, des forêts aléatoires et la recalibration lorsque la sortie en aval est probabiliste. Vérifiez si les fonctionnalités découvertes aident en dehors de CatBoost.
- Ajouter une référence par embedding. Un embedding transforme une note en quelques centaines de nombres sans question associée :
sentence-transformers/all-MiniLM-L6-v2s’exécute localement,text-embedding-3-smalld’OpenAI est un appel hébergé. Ajoutez-en un aux colonnes découvertes et mesurez s’il apporte des informations que celles-ci ne couvrent pas. - Adapter la validation au déploiement. Utilisez des partitions chronologiques pour prédire l’avenir, des partitions groupées lorsque des lignes apparentées doivent rester ensemble, et conservez un ensemble de test final intact, non utilisé ni pour la découverte de fonctionnalités ni pour la sélection de modèle.
- Arrêter sur un plateau. Terminez la boucle lorsque la RMSE croisée ne s’améliore plus pendant un nombre fixe de tours, ou lorsqu’elle atteint un budget de questions ou de requêtes.
- Exécuter une recherche plus longue en mode Goal d’un agent. Donnez-lui une métrique explicite, un budget et une règle d’arrêt, puis laissez-le proposer, évaluer et affiner sur davantage de tours.
- Vérifier la stabilité. Répétez la découverte sur différentes graines ou tranches de données et conservez les questions qui restent utiles, plutôt que celles dont l’importance repose sur une seule partition.
Ouvrez-le dans le terrain de jeu
Ce lien de partage contient une note de dégustation ainsi que toutes les questions que la boucle a finalement générées.
playground_link = make_playground_link(
NOTES[0], feature_questions(accepted), models=[TYPESAFE_MODEL]
)
display(
Markdown(
f"🔗 [Open the note + questions in the TypeSafe playground]({playground_link})"
)
)Ouvrir la note + les questions dans le playground TypeSafe →