When to Trust the Answer: Question-Aligned Semantic Nearest Neighbor Entropy for Safer Surgical VQA
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910159983345664 |
|---|---|
| author | Carlini, Luca Pierantozzi, Dennis Drago, Mauro Orazio Lena, Chiara Hassan, Cesare De Momi, Elena Stoyanov, Danail Bano, Sophia Hoque, Mobarak I. |
| author_facet | Carlini, Luca Pierantozzi, Dennis Drago, Mauro Orazio Lena, Chiara Hassan, Cesare De Momi, Elena Stoyanov, Danail Bano, Sophia Hoque, Mobarak I. |
| contents | Safety and reliability are critical for deploying visual question answering (VQA) systems in surgery, where incorrect or ambiguous responses can cause patient harm. A key limitation of existing uncertainty estimation methods, such as Semantic Nearest Neighbor Entropy (SNNE), is that they do not explicitly account for the conditioning question. As a result, they may assign high confidence to answers that are semantically consistent yet misaligned with the clinical question, especially under variation in question phrasing. We propose Question-Aligned Semantic Nearest Neighbor Entropy (QA-SNNE), a black-box uncertainty estimator that incorporates question-answer alignment into semantic entropy through bilateral gating. QA-SNNE measures uncertainty by weighting pairwise semantic similarities among sampled answers according to their relevance to the question, using embedding-based, entailment-based, or cross-encoder alignment strategies. To assess robustness to language variation, we construct an out-of-template rephrased version of a benchmark surgical VQA dataset, where only the question wording is modified while images and ground-truth answers remain unchanged. We evaluate QA-SNNE on five VQA models across two benchmark surgical VQA datasets in both zero-shot and parameter-efficient fine-tuned (PEFT) settings, including out-of-template questions. QA-SNNE improves AUROC on EndoVis18-VQA for two of three zero-shot models in-template (e.g., +15% for Llama3.2 and +21% for Qwen2.5) and achieves up to +8% AUROC improvement under out-of-template rephrasing, with mixed results on external validation. Overall, QA-SNNE provides a practical, model-agnostic safeguard for surgical VQA by linking semantic uncertainty to question relevance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_01458 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | When to Trust the Answer: Question-Aligned Semantic Nearest Neighbor Entropy for Safer Surgical VQA Carlini, Luca Pierantozzi, Dennis Drago, Mauro Orazio Lena, Chiara Hassan, Cesare De Momi, Elena Stoyanov, Danail Bano, Sophia Hoque, Mobarak I. Computer Vision and Pattern Recognition Artificial Intelligence Safety and reliability are critical for deploying visual question answering (VQA) systems in surgery, where incorrect or ambiguous responses can cause patient harm. A key limitation of existing uncertainty estimation methods, such as Semantic Nearest Neighbor Entropy (SNNE), is that they do not explicitly account for the conditioning question. As a result, they may assign high confidence to answers that are semantically consistent yet misaligned with the clinical question, especially under variation in question phrasing. We propose Question-Aligned Semantic Nearest Neighbor Entropy (QA-SNNE), a black-box uncertainty estimator that incorporates question-answer alignment into semantic entropy through bilateral gating. QA-SNNE measures uncertainty by weighting pairwise semantic similarities among sampled answers according to their relevance to the question, using embedding-based, entailment-based, or cross-encoder alignment strategies. To assess robustness to language variation, we construct an out-of-template rephrased version of a benchmark surgical VQA dataset, where only the question wording is modified while images and ground-truth answers remain unchanged. We evaluate QA-SNNE on five VQA models across two benchmark surgical VQA datasets in both zero-shot and parameter-efficient fine-tuned (PEFT) settings, including out-of-template questions. QA-SNNE improves AUROC on EndoVis18-VQA for two of three zero-shot models in-template (e.g., +15% for Llama3.2 and +21% for Qwen2.5) and achieves up to +8% AUROC improvement under out-of-template rephrasing, with mixed results on external validation. Overall, QA-SNNE provides a practical, model-agnostic safeguard for surgical VQA by linking semantic uncertainty to question relevance. |
| title | When to Trust the Answer: Question-Aligned Semantic Nearest Neighbor Entropy for Safer Surgical VQA |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2511.01458 |