Fine-Tuning GPT-3.5-Turbo for Automatic Feedback Generation
Bibliographic record
Abstract
Scaling up the delivery of effective feedback remains an open challenge in education. Existing automatic feedback generation (AFG) methods fall short in providing feedback highly tailored to tasks, students, and instructors' preferences, simultaneously. Recent evidence suggests that Large Language Models (LLMs), with their ability to follow instructions and generate text, could address this limitation. Existing studies have generated feedback using GPT models in their ready-to-use Chat version, using almost exclusively prompting strategies to direct the model towards the desired output. Results are largely positive; however, space for improvement remains. For the first time, the present study reports observations and results from fine-tuning GPT-3.5-turbo for AFG for open-ended situational judgment questions from the high-stakes test Casper. The LLM was fine-tuned using a small set of hand-written feedback examples, and independent judges and text experts evaluated model performance using a rubric based on qualities of effective feedback identified in the literature. Moreover, a survey study measured users' satisfaction with automatic feedback. Results show that, although not perfect, the fine-tuned model generated outputs largely aligned with the desired qualities and often aligned with the given guidelines, satisfying the majority of users. The strengths and weaknesses of our model are discussed, and directions for future research are suggested.
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 | 0.005 | 0.038 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.007 | 0.004 |
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".