TSPPT: Two-Stage Prompt Pre-Train to Promote Few-Shot Learning Performance
Bibliographic record
Abstract
The Pretrained-Language Model (PLM) has achieved dominance in the field of Natural Language Processing (NLP), and prompt learning further enhances its impact by aligning the pre-training tasks of the language model with the downstream tasks. However, comparing with traditional fine-tune, prompt learning has some disadvantages such as poor absolute accuracy, low training efficiency and poor robustness, especially in the case of small parameters of the language model itself or insufficient training data. A large number of studies have shown that the main defect of Prompt learning (PL) at the present stage is that the quality of Prompt itself plays an important role in the performance of the model, and the existing initialization method of prompt is often not optimal. Therefore, we propose Two-Stage Prompt Pre-Train (TSPPT): using the special pre-training tasks, obtained by constructing or reforming raw texts and downstream tasks, to pre-train two sub-prompt, Task-oriented sub-Prompt (TSP) and Universal Sub-Prompt (USP), in two advanced stages respectively. By concatenating USP and TSP as the prompt initialization for language model to prompt-tuning on downstream tasks, TSPPT promotes overall performance, such as robustness, accuracy, and generalization. Experiments have shown that TSPPT can achieve or even exceed the performance of traditional fine-tuning while retaining the advantage freezing language model parameters and tuning few parameters only.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.006 | 0.003 |
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 source (direct Gemma or distilled Codex), 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".