REVIEW 12 cited by
Efficient Few-Shot Learning Without Prompts
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Recent few-shot methods, such as parameter-efficient fine-tuning (PEFT) and pattern exploiting training (PET), have achieved impressive results in label-scarce settings. However, they are difficult to employ since they are subject to high variability from manually crafted prompts, and typically require billion-parameter language models to achieve high accuracy. To address these shortcomings, we propose SetFit (Sentence Transformer Fine-tuning), an efficient and prompt-free framework for few-shot fine-tuning of Sentence Transformers (ST). SetFit works by first fine-tuning a pretrained ST on a small number of text pairs, in a contrastive Siamese manner. The resulting model is then used to generate rich text embeddings, which are used to train a classification head. This simple framework requires no prompts or verbalizers, and achieves high accuracy with orders of magnitude less parameters than existing techniques. Our experiments show that SetFit obtains comparable results with PEFT and PET techniques, while being an order of magnitude faster to train. We also show that SetFit can be applied in multilingual settings by simply switching the ST body. Our code is available at https://github.com/huggingface/setfit and our datasets at https://huggingface.co/setfit .
Forward citations
Cited by 12 Pith papers
-
How Much Does a Reasoning Summary Reveal? An Observability Ladder for Large Language Models
With the prompt visible, LLM-written reasoning summaries add almost no correctness signal for linear readers, while full traces still add signal; monitorability is a joint property of display and reader.
-
Test-Time Scaling via Error Localization
TTEL uses feedback-induced token probability drops to localize the first error in a failed reasoning trace and branch a new generation from that prefix, improving pass@k per token on coding and math benchmarks.
-
From Regulation to Requirements: An Automated Requirement Derivation and Explanation Pipeline
A two-stage GPT-5 pipeline identifies requirement-bearing regulatory clauses (F1 0.82/0.78) and derives traceable software requirements with explanations, evaluated on full GDPR and EU AI Act texts plus a 25-person us...
-
Few-Shot Query Intent Detection via Relation-Aware Prompt Learning
SAID pretrains language models with query-query and query-answer relation-aware soft prompts, then transfers them via intent-specific prompts for few-shot intent detection, reporting up to 27% relative accuracy gains.
-
ConspirED: A Dataset for Cognitive Traits of Conspiracy Theories and Large Language Model Safety
A new dataset annotates conspiracy texts with six cognitive traits, and experiments show LLMs reproduce conspiracy reasoning more readily than they deflect it.
-
On The Impact of Merge Request Deviations on Code Review Practices
A taxonomy of seven non-review merge request types, a few-shot classifier that detects them, and evidence that excluding them changes code review analytics.
-
Unveiling Dual Quality in Product Reviews: An NLP-Based Approach
A new Polish dataset and model evaluation show that fine-tuned Polish transformers detect dual-quality product reviews with precision close to GPT-4o.
-
ReqBrain: Task-Specific Instruction Tuning of LLMs for AI-Assisted Requirements Generation
ReqBrain, a LoRA-fine-tuned Zephyr-7b-beta model, produces software requirements that human evaluators could not reliably tell apart from human-authored ones, with automatic metrics favoring it over untuned ChatGPT-4o.
-
A Longitudinal Analysis of Public Discourse on AI Ethics in Education Using Twitter Data
US Twitter discourse on AI ethics in education (2019–2024) is ~82% positive, pragmatic about classroom use, and increasingly focused on academic integrity after ChatGPT.
-
The Narrative Construction of Generative AI Efficacy by the Media: A Case Study of the Role of ChatGPT in Higher Education
U.S. news media predominantly frame ChatGPT's role in higher education as positive, with negative coverage concentrating on entry-level jobs and college admissions.
-
Self-Anchored Attention Model for Sample-Efficient Classification of Prosocial Text Chat
A self-anchored attention classifier using the whole training set as anchor features achieves 0.836 AUC for prosocial chat in Call of Duty: Modern Warfare II, 7.9% above the best benchmark.
-
Evaluating the Performance and Efficiency of Sentence-BERT for Code Comment Classification
On the NLBSE 2025 code comment classification task, small Sentence-BERT models with tuned classifiers improve F1 by +0.0346 with only 1.4x runtime and 2.1x GFLOPS overhead.
Discussion (0). Sign in to comment.