Research map / Vision-language and multimodal / Visual question answering
Predict Before You Explore: Predictive Planning with Specialized Memory for Embodied Question Answering
Bowen Yuan, Sisi You, Bing-Kun Bao
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 predictive hierarchical planner to reactive baseline (ablation) | Pred-EQA · A-EQA, Qwen3-VL 8B | LLM-SPL: 39.5 -> 45.1 | Improved |
| add predictive planner to reactive baseline (ablation) | Pred-EQA · A-EQA, Qwen2.5-VL 7B | LLM-Match: 40.8 -> 44.3 | Improved |
| add specialized dual memory only (ablation) | Pred-EQA · A-EQA, Qwen2.5-VL 7B | LLM-SPL: 24.6 -> 28.7 | Improved |
| combine predictive planner with specialized memory (ablation) | Pred-EQA · A-EQA, Qwen2.5-VL 7B | LLM-SPL: 24.6 -> 37.8 | Improved |
| combine predictive planner with specialized memory (ablation) | Pred-EQA · A-EQA, Qwen3-VL 8B | LLM-Match: 45.7 -> 53.3 | Improved |
| prompt reactive executor to focus on long-term planning (ablation) | Qwen3-VL 8B baseline, no memory · A-EQA | LLM-Match: 45.7 -> 44.0 | Hurt |
| use TODO-list planning instead of predictive planning (ablation) | Qwen3-VL 8B, no memory · A-EQA | LLM-SPL: 39.5 -> 39.9 | No clear effect |
| predictive plan versus TODO-list plan (ablation) | Qwen3-VL 8B, no memory · A-EQA | LLM-Match: 47.7 -> 50.8 | Improved |
| predictive plan versus TODO-list plan with specialized memory (ablation) | Qwen3-VL 8B · A-EQA | LLM-SPL: 43.9 -> 48.5 | Improved |
| replace specialized memory with 3D-Mem under reactive planning (ablation) | Qwen3-VL 8B baseline · A-EQA | LLM-Match: 45.7 -> 48.8 | Improved |
| specialized memory versus Manager* with predictive planning (ablation) | Pred-EQA Qwen3-VL 8B · A-EQA | LLM-Match: 53.1 -> 53.3 | Improved |
| scale VLM from Qwen3-VL 4B to 32B (ablation) | Pred-EQA · A-EQA | LLM-Match: 50.8 -> 56.3 | Improved |
Closest papers by meaning
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.91)
- Exploratory Retrieval-Augmented Planning For Continual Embodied Instruction Following (NeurIPS 2024 · similarity 0.89)
- Embodied Task Planning via Graph-Informed Action Generation with Large Language Models (ICML 2026 · similarity 0.88)
- Seeing, Listening, Remembering, and Reasoning: A Multimodal Agent with Long-Term Memory (ICLR 2026 · similarity 0.88)
- Embodied Agents Meet Personalization: Investigating Challenges and Solutions Through the Lens of Memory Utilization (ICLR 2026 · similarity 0.87)
- More Than Meets the Eye? Uncovering the Reasoning-Planning Disconnect in Training Vision-Language Driving Models (NeurIPS 2026 · 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.
