Evaluating Zero-Shot and Few-Shot Learning on the Turkish Massive Multitask Language Understanding Dataset
Bibliographic record
Abstract
This study investigates the effectiveness of the zeroshot and few-shot learning paradigms on the Turkish Massive Multitask Language Understanding (MMLU) dataset, a translated version of the MMLU benchmark. The Turkish MMLU dataset, comprising 293,468 multi-disciplinary multiple-choice questions, is the largest dataset for Turkish academic and professional exams, covering subjects such as STEM, humanities, and social sciences. The primary objective of this research is to evaluate the strengths and limitations of the zero-shot and fewshot learning paradigms in Turkish natural language processing (NLP) tasks, providing insights into effective methodologies for knowledge-intensive tasks in Turkish. The study addresses the lack of comprehensive evaluations of learning paradigms in Turkish NLP, aiming to optimize resource allocation and advance the development of state-of-the-art language models for Turkish. The proposed solution involves a structured experimentation framework that compares the zero-shot and few-shot learning paradigms. Zero-shot learning leverages pre-trained language models to classify text into unseen categories using task descriptions, while few-shot learning utilizes a small number of labeled examples to generalize to new queries. This research contributes to the field by providing a robust framework for evaluating learning paradigms and advancing Turkish language technology. It evaluates the performance of multiple models, including Turkish BERT, mDeBERTa-v3, and Owen/Owen2-0.5B, on the Turkish MMLU dataset. The results reported in this paper indicate that few-shot learning consistently outperforms zero-shot learning.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.002 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".