afk-setfit-camembert-base-v1

French news headline classifier for AFK.live, fine-tuned with SetFit.

Quick Start

from setfit import SetFitModel

model = SetFitModel.from_pretrained("afk-live/afk-setfit-camembert-base-v1")
model.predict(["Macron annonce une réforme majeure des retraites"])
# => ["Politique France"]

Model Details

Attribute	Value
Base Model	almanach/camembert-base
Accuracy	80.57%
Mean F1	0.7962
Training Time	812s
Dataset	afk-live/afk-news-fr-classification-202601

Categories (12 labels)

ID	Label	Description	Live
1	Politique FR, EU, Québec	Domestic francophone politics: France, Belgium, Switzerland, Quebec, EU institutions	Yes
2	Géopolitique mondiale	International relations, conflicts, diplomacy outside francophone sphere	Yes
3	Economie & Marchés & Business Tech	Pure finance, macroeconomics, non-tech corporations, banking, markets	Yes
4	Société		Yes
5	Science & Environnement	Research discoveries, climate, biodiversity, health research (non-clickbait)	Yes
6	Culture	Cultural works and events: cinema, music, books, theater, visual arts, video games	Yes
7	Sport	All sports news (competitions, transfers, results)	Yes
8	Faits divers & Actualité locale	Accidents, local news without national relevance	No
10	Gadgets, pub, articles sponsorisés	Mode, cooking, travel, wellness, fashion, product reviews, deals	No
11	Météo	Weather forecasts, daily weather reports	No
12	Rediffusion, replay, programme TV	TV schedules, replay announcements, program grids	No
13	Code & Infrastructure	Developer-focused technical content: programming, ML engineering, DevOps, software architecture	Yes

Note: "Live" indicates whether the category is displayed on AFK.live (Yes) or filtered out (No).

Training Details

• Train samples: 720 (60 per class)
• Test samples: 211
• Hyperparameters: batch_size=(32,16), epochs=(1,16), end_to_end=True, sampling=unique

About AFK.live

AFK.live is a French news aggregation project that creates trends from journalistic work.

Last updated: 2026-01-17