Fine-tuning Qwen3-VL-32B-Instruct on a curated set of 13k fracture images yields a specialist model achieving 0.92 precision on morphology recognition, outperforming the base model and several proprietary VLMs on a 100-image manual benchmark.
https://developers.openai.com/api/docs/models/gpt-5.2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.mtrl-sci 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Fine-tuning a vision-language model for fracture-surface morphology recognition
Fine-tuning Qwen3-VL-32B-Instruct on a curated set of 13k fracture images yields a specialist model achieving 0.92 precision on morphology recognition, outperforming the base model and several proprietary VLMs on a 100-image manual benchmark.