Research map / Vision-language and multimodal / Multimodal LLMs
DoCoG: Mask-based Multi-Type Grounded Chain-of-Thought for Document QA
Sai Madhusudan Gunda, Jyothi Jinka, Hrithik Sagar, Aryan Jain, Venkata Venna, Anirudh Srinivasan, SANTOSH RAVI KIRAN SARVADEVABHATLA
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 grounding of final answer (ablation) | DoCoG · DoCoG-Bench | AQ: 78.1 -> 81.5 | Improved |
| extend grounding to intermediate CoT steps (ablation) | DoCoG · DoCoG-Bench | AQ: 81.5 -> 84.1 | Improved |
| add DPO stage after SFT (ablation) | DoCoG · DoCoG-Bench | AQ: 84.1 -> 85.8 | Improved |
| add step-wise CoT (ablation) | DoCoG · DoCoG-Bench | AQ: 73.3 -> 85.8 | Improved |
| add Grounding Interaction Module (GIM) (ablation) | DoCoG · DoCoG-Bench | GQfa: 69.6 -> 81.9 | Improved |
| add image cross-attention in GIM (ablation) | DoCoG · DoCoG-Bench | GQall: 73.5 -> 78.2 | Improved |
| add anchor loss during DPO (ablation) | DoCoG · DoCoG-Bench | GQfa: 41.7 -> 81.9 | Improved |
| add anchor loss during DPO (ablation) | DoCoG · MMDocBench | IoU: 40.3 -> 70.5 | Improved |
| use full fine-tuning instead of LoRA (ablation) | DoCoG · DoCoG-Bench | GQfa: 65.5 -> 81.9 | Improved |
| fine-tune Qwen3-VL on DoCoG-QA | Qwen3-VL-8B · DoCoG-Bench | GQfa: 5.5 -> 61.9 | Improved |
| use DoCoG architecture (GIM, mask supervision) over fine-tuned Qwen3-VL | Qwen3-VL FT · DoCoG-Bench | AQ: 77.5 -> 85.8 | Improved |
Closest papers by meaning
- M3Grounder: Mask-Based Multi-Span and Multi-Granular Grounding for Document QA (CVPR 2026 · similarity 0.93)
- DocVAL: Validated Chain-of-Thought Distillation for Grounded Document VQA (ICML 2026 · similarity 0.90)
- Benchmarking Retrieval-Augmented Multimomal Generation for Document Question Answering (NeurIPS 2025 · similarity 0.89)
- DocScope: Benchmarking Verifiable Reasoning for Trustworthy Long-Document Understanding (NeurIPS 2026 · similarity 0.88)
- Marten: Visual Question Answering with Mask Generation for Multi-modal Document Understanding (CVPR 2025 · similarity 0.88)
- GRAM: Global Reasoning for Multi-Page VQA (CVPR 2024 · similarity 0.87)
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.
