Research map / Vision-language and multimodal / Visual question answering
OIDA-QA: A Multimodal Benchmark for Analyzing the Opioid Industry Documents Archive
Xuan Shen, Brian Wingenroth, Zichao Wang, Jason Kuen, Wanrong Zhu, Ruiyi Zhang, Yiwei Wang, Lichun Ma, Anqi Liu, Hongfu Liu, Tong Sun, Kevin S. Hawkins and 3 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Page Finder when no ground-truth window is available (ablation) | Mistral-7B OIDA-QA model · OIDA-QA test | BLEU-4: 38.5% -> 48.6% | Improved |
| add Page Finder when no ground-truth window is available (ablation) | Mistral-7B OIDA-QA model · OIDA-QA page grounding | page generation rate: 68.7% -> 88.2% | Improved |
| add content reiteration (ablation) | window-1 model · OIDA-QA test | BLEU-4: 49.7% -> 51.0% | Improved |
| add Page Finder on window-1 model without reiteration (ablation) | window-1 model · OIDA-QA test | BLEU-4: 49.7% -> 48.1% | Hurt |
| add Page Finder and content reiteration together (ablation) | window-1 model · OIDA-QA test | BLEU-4: 49.7% -> 53.7% | Improved |
| add Page Finder (ablation) | window-3 model · OIDA-QA test | BLEU-4: 48.7% -> 52.0% | Improved |
| add content reiteration alone (ablation) | window-3 model · OIDA-QA page grounding | page generation rate: 80.4% -> 79.2% | No clear effect |
| increase context window from 1 to max (ablation) | base model without extras · OIDA-QA page grounding | page generation rate: 69.7% -> 80.8% | Improved |
| train without any context window (ablation) | Mistral-7B OIDA-QA model · OIDA-QA test | BLEU-4: 38.5% vs 49.7% (window 1) | Hurt |
| add content reiteration to small window (ablation) | window-1 model · OIDA-QA page grounding | page generation rate: 69.7% -> 73.9% | Improved |
Closest papers by meaning
- Characterizing Information Seeking Events in Health-Related Social Discourse (AAAI 2024 · similarity 0.84)
- Benchmarking Retrieval-Augmented Multimomal Generation for Document Question Answering (NeurIPS 2025 · similarity 0.83)
- SPIQA: A Dataset for Multimodal Question Answering on Scientific Papers (NeurIPS 2024 · similarity 0.82)
- EHRNoteQA: An LLM Benchmark for Real-World Clinical Practice Using Discharge Summaries (NeurIPS 2024 · similarity 0.82)
- EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries (NeurIPS 2026 · similarity 0.82)
- CLIPSyntel: CLIP and LLM Synergy for Multimodal Question Summarization in Healthcare (AAAI 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
