Hyperparameter	Value
LANGUAGES	en-es-it
LR	2e-05
EPOCHS	3
MAX_LENGTH	256
USE_BIO	False
USE_LANG_TOKEN	False
GATED_BIO	False
FOCAL_LOSS	True
FOCAL_GAMMA	1.5
USE_SAMPLER	True
R_DROP	True
R_KL_ALPHA	0.5
TEXT_NORMALIZE	True

Dev set results (summary)

Metric	Value
f1_macro_dev_0.5	0.6822920086077982
f1_weighted_dev_0.5	0.8107834514586362
accuracy_dev_0.5	0.7839643652561247
f1_macro_dev_best_global	0.7327145250903959
f1_weighted_dev_best_global	0.8744347857513664
accuracy_dev_best_global	0.8797327394209354
f1_macro_dev_best_by_lang	0.750740192450037
f1_weighted_dev_best_by_lang	0.8789340239598152
accuracy_dev_best_by_lang	0.8797327394209354
default_threshold	0.5
best_threshold_global	0.8
thresholds_by_lang	{"en": 0.8, "it": 0.65, "es": 0.8}

Thresholds

Default: 0.5
Best global: 0.8
Best by language: { "en": 0.8, "it": 0.65, "es": 0.8 }

Detailed evaluation

Classification report @ 0.5

text

              precision    recall  f1-score   support

 no-recl (0)     0.9528    0.7870    0.8620       385
    recl (1)     0.3740    0.7656    0.5026        64

    accuracy                         0.7840       449
   macro avg     0.6634    0.7763    0.6823       449
weighted avg     0.8703    0.7840    0.8108       449

Classification report @ best global threshold (t=0.80)

text

              precision    recall  f1-score   support

 no-recl (0)     0.9169    0.9455    0.9309       385
    recl (1)     0.5962    0.4844    0.5345        64

    accuracy                         0.8797       449
   macro avg     0.7565    0.7149    0.7327       449
weighted avg     0.8712    0.8797    0.8744       449

Classification report @ best per-language thresholds

text

              precision    recall  f1-score   support

 no-recl (0)     0.9276    0.9325    0.9301       385
    recl (1)     0.5806    0.5625    0.5714        64

    accuracy                         0.8797       449
   macro avg     0.7541    0.7475    0.7507       449
weighted avg     0.8782    0.8797    0.8789       449

Per-language metrics (at best-by-lang)

lang	n	acc	f1_macro	f1_weighted	prec_macro	rec_macro	prec_weighted	rec_weighted
en	154	0.8831	0.5594	0.8733	0.5732	0.5521	0.8649	0.8831
it	163	0.8834	0.8084	0.8827	0.8124	0.8046	0.8821	0.8834
es	132	0.8712	0.7639	0.8748	0.7505	0.7804	0.8796	0.8712

Data

Train/Dev: private multilingual splits with ~15% stratified Dev (by (lang,label)).
Source: merged EN/IT/ES data with bios retained (ignored if unused by model).

Usage

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoConfig
import torch, numpy as np

repo = "SimoneAstarita/october-finetuning-more-variables-sweep-20251012-201341-t07"
tok = AutoTokenizer.from_pretrained(repo)
cfg = AutoConfig.from_pretrained(repo)
model = AutoModelForSequenceClassification.from_pretrained(repo)

texts = ["example text ..."]
langs = ["en"]

mode = "best_global"  # or "0.5", "by_lang"

enc = tok(texts, truncation=True, padding=True, max_length=256, return_tensors="pt")
with torch.no_grad():
    logits = model(**enc).logits
probs = torch.softmax(logits, dim=-1)[:, 1].cpu().numpy()

if mode == "0.5":
    th = 0.5
    preds = (probs >= th).astype(int)
elif mode == "best_global":
    th = getattr(cfg, "best_threshold_global", 0.5)
    preds = (probs >= th).astype(int)
elif mode == "by_lang":
    th_by_lang = getattr(cfg, "thresholds_by_lang", {})
    preds = np.zeros_like(probs, dtype=int)
    for lg in np.unique(langs):
        t = th_by_lang.get(lg, getattr(cfg, "best_threshold_global", 0.5))
        preds[np.array(langs) == lg] = (probs[np.array(langs) == lg] >= t).astype(int)
print(list(zip(texts, preds, probs)))

Additional files

reports.json: all metrics (macro/weighted/accuracy) for @0.5, @best_global, and @best_by_lang. config.json: stores thresholds: default_threshold, best_threshold_global, thresholds_by_lang. postprocessing.json: duplicate threshold info for external tools.