Research map / Vision-language and multimodal / Multimodal LLMs
Visual Intention Grounding for Egocentric Assistants
Pengzhan Sun, Junbin Xiao, Tze Ho Elden Tse, Yicong Li, Arjun Akula, Angela Yao
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 |
|---|---|---|---|
| use RoG (reason then ground) instead of naive SFT | MiniGPT-v2 · EgoIntention (overall) | Precision@0.5: 43.4 -> 47.3 | Improved |
| use RoG instead of naive SFT | Qwen-VL · EgoIntention (overall) | Precision@0.5: 29.1 -> 33.6 | Improved |
| use RoG SFT instead of naive SFT | MiniGPT-v2 · RefCOCO val | Precision@0.5: 86.6 -> 87.8 | Improved |
| use RoG SFT instead of naive SFT | Qwen-VL · RefCOCO val | Precision@0.5: 89.5 -> 89.3 | No clear effect |
| run reasoning (GPT-4) before detection (GroundingDINO) | two-stage hybrid baseline · EgoIntention (context split) | Precision@0.5: 21.1 -> 46.6 | Improved |
| run reasoning before detection | two-stage hybrid baseline · EgoIntention (uncommon split) | Precision@0.5: 14.6 -> 23.6 | Improved |
| fine-tune on EgoIntention only without RefCOCO/+/g (ablation) | MiniGPT-v2 · RefCOCO val | Precision@0.5: 87.4 -> 66.5 | Hurt |
| add RefCOCO/+/g to EgoIntention in SFT (ablation) | MiniGPT-v2 · EgoIntention (context) | Precision@0.5: 42.8 -> 45.9 | Improved |
| add RefCOCO/+/g to EgoIntention in SFT (ablation) | MiniGPT-v2 · EgoIntention (uncommon) | Precision@0.5: 39.2 -> 40.8 | Improved |
| add auto-generated RefCOCOInt/+/g intention data (ablation) | naive SFT MiniGPT-v2 · EgoIntention (overall) | Precision@0.5: 43.3 -> 43.4 | No clear effect |
| use RoG SFT instead of naive SFT on explicit object queries | MiniGPT-v2 · EgoIntention explicit object | Precision@0.5: 51.3 -> 52.2 | Improved |
| add EgoIntention uncommon split to training (ablation) | RoG SFT MiniGPT-v2 · EgoIntention uncommon | Precision@0.5: 33.7 -> 44.7 | Improved |
Closest papers by meaning
- ANNEXE: Unified Analyzing, Answering, and Pixel Grounding for Egocentric Interaction (CVPR 2025 · similarity 0.89)
- In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting (NeurIPS 2025 · similarity 0.89)
- EARL: Towards a Unified Analysis-Guided Reinforcement Learning Framework for Egocentric Interaction Reasoning and Pixel Grounding (ICML 2026 · similarity 0.88)
- DenseGrounding: Improving Dense Language-Vision Semantics for Ego-centric 3D Visual Grounding (ICLR 2025 · similarity 0.88)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.88)
- Is `Right' Right? Enhancing Object Orientation Understanding in Multimodal Large Language Models through Egocentric Instruction Tuning (CVPR 2025 · similarity 0.88)
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.
