SetFit v1.2 released: 8 examples per class rivals RoBERTa-Large fine-tuned on 3k
tomaarsen · x · 2026-09-04
Hugging Face released SetFit v1.2.0, a prompt-free few-shot text classification framework: it builds positive/negative pairs from a handful of labeled sentences, contrastively fine-tunes a Sentence Transformer, and fits a classifier on the embeddings — no prompts or LLMs needed. The original paper shows 8 examples per class rivals RoBERTa-Large fine-tuned on 3k examples, training an order of magnitude faster than PEFT/PET. v1.2 adds support for transformers v5, Sentence Transformers v6, and huggingfacehub v1.
More from Research
- Three reasons AI is not conscious and may never be, per Nature essay — anilkseth · 2026-10-21
- Aviation tracks every part's provenance — agents can reason over that graph — colinmcnamara · 2026-10-21
- SFT details: instruction/reasoning/code/math mix, synthetic + human traces, no epoching — TimDarcet · 2026-10-21
- Mid-training recipe disclosed: 32k→131k context extension, dropping logit distillation for tokens — TimDarcet · 2026-10-21
- Mathematician Daniel Litt concedes his AI bet, but says AI won't solve math — ziv_ravid · 2026-10-21
- AI drug discovery reality check: 'give me everything that binds,' then filter for toxicity — burny_tech · 2026-10-21