Can Large Vision-Language Models Understand Construction Safety? A Novel Benchmark Using Construction Safety Posters
Bibliographic record
Abstract
Manual detection of safety issues onsite is error-prone, labor-intensive, and costly. Advancements in Artificial intelligence (AI) ushered in the era of automated safety detection using images and videos collected onsite. Recent developments of large vision-language models (VLMs) proved AI’s ability to understand common concepts, which allows these pre-trained models to be swiftly applied to new tasks without further training. The natural question to ask is whether these large models trained on data scraped from the internet understand concepts related to construction safety. Evaluating this capability is critical before deploying pre-trained models on real construction sites, especially in zero-shot settings (i.e., directly deploying models without training). This study marks the first attempt at evaluating AI’s capability of understanding safety concepts in zero-shot settings by deploying a state-of-the-art pre-trained VLM to match 443 construction safety posters collected from the internet (i.e., visual inputs) with detailed descriptions of the posters written manually by the researchers (i.e., natural language inputs). Results reveal a significant 93% accuracy in associating posters with correct descriptions. This work proposes a new task of benchmarking large models’ capability of understanding complex concepts onsite without training, paving the way for large-scale evaluation and deployment of these models onsite.
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 | 0.003 | 0.013 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
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".