Research map / Vision-language and multimodal / CLIP and zero-shot
Plug-and-Play Clarifier: A Zero-Shot Multimodal Framework for Egocentric Intent Disambiguation
Sicheng Yang, Yukai Huang, Weitong Cai, Shitong Sun, You He, Jiankang Deng, Hang Zhang, Jifei Song, Zhensong Zhang
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 iterative dialogue-driven text clarifier to a small LLM | Qwen2.5-7B · CLAMBER | vagueness judgement accuracy (%): 24.4 -> 53.0 | Improved |
| add iterative dialogue-driven text clarifier to a small LLM | Llama-3.1-8B · CLAMBER | vagueness judgement accuracy (%): 25.9 -> 52.9 | Improved |
| add text clarifier to a large LLM | Llama-3.1-405B · CLAMBER | vagueness judgement accuracy (%): 54.9 -> 60.1 | Improved |
| add vision clarifier (open-set detector plus geometric quality checks) | GPT-4o · VRA-Ego visual ambiguity | strict recover rate (%): 40.6 -> 61.4 | Improved |
| add vision clarifier | Qwen2.5-VL · VRA-Ego visual ambiguity | target identification accuracy (%): 86.6 -> 91.2 | Improved |
| add cross-modal 3D pointing ray-casting with context-aware crop | GPT-4o · VRA-Ego referential ambiguity | semantic answer recover rate (%): 65.2 -> 69.0 | Improved |
| add cross-modal pointing clarifier | Qwen2.5-VL · VRA-Ego referential ambiguity | semantic answer recover rate (%): 57.8 -> 64.4 | Improved |
| swap default open-set detector for Florence-2 (ablation) | Vision clarifier · VRA-Ego visual ambiguity | accuracy / recover rate: about 1% accuracy, 3% recover rate drop | Hurt |
| swap default detector for YOLOE or YOLO-World (ablation) | Vision clarifier · VRA-Ego visual ambiguity | accuracy / recover rate: about 2% accuracy, 5% recover rate drop | Hurt |
| replace custom fingertip detector with MediaPipe (ablation) | Cross-modal clarifier · VRA-Ego referential ambiguity | pointing success accuracy: 15% decrease | Hurt |
| use context-aware cropping instead of point-based segmentation or full image (ablation) | Cross-modal clarifier · VRA-Ego referential ambiguity | semantic answer recover rate | Improved |
Closest papers by meaning
- EARL: Towards a Unified Analysis-Guided Reinforcement Learning Framework for Egocentric Interaction Reasoning and Pixel Grounding (ICML 2026 · similarity 0.88)
- In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting (NeurIPS 2025 · similarity 0.88)
- EgoProx: Evaluating MLLMs on Egocentric 3D Proximity Reasoning Across a Cognitive Hierarchy (CVPR 2026 · similarity 0.87)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.86)
- Is `Right' Right? Enhancing Object Orientation Understanding in Multimodal Large Language Models through Egocentric Instruction Tuning (CVPR 2025 · similarity 0.86)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.86)
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.
