Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Factorizing Perception and Policy for Interactive Instruction Following
Kunal Pratap Singh, Suvaansh Bhambri, Byeonghwi Kim, Roozbeh Mottaghi, Jonghyun Choi
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 factorized perception and policy streams (ablation) | MOCA · ALFRED val seen | task success rate: 22.32 -> 25.85 | Improved |
| add factorization when data augmentation is absent (ablation) | MOCA · ALFRED val seen | task success rate: 12.56 -> 15.85 | Improved |
| add data augmentation (ablation) | MOCA · ALFRED val seen | task success rate: 15.85 -> 25.85 | Improved |
| add language-guided dynamic filters (ablation) | MOCA · ALFRED val seen | task success rate: 14.63 -> 15.85 | Improved |
| add language-guided dynamic filters (ablation) | MOCA · ALFRED val unseen | task success rate: 2.19 -> 2.92 | Improved |
| add object-centric localisation (ablation) | MOCA · ALFRED val seen | task success rate: 3.90 -> 15.85 | Improved |
| add object-centric localisation with factorization (ablation) | MOCA · ALFRED val seen | task success rate: 11.71 -> 14.63 | Improved |
| add instance association (ablation) | MOCA · ALFRED val seen | task success rate: 23.66 -> 25.85 | Improved |
| add obstruction evasion (ablation) | MOCA · ALFRED val seen | task success rate: 20.00 -> 25.85 | Improved |
| feed goal and instructions concatenated to IPM (ablation) | MOCA · ALFRED val seen | task success rate: 25.85 -> 29.76 | Mixed |
| feed goal and instructions concatenated to APM (ablation) | MOCA · ALFRED val unseen | task success rate: 5.36 -> 5.36 | No clear effect |
| feed goal and instructions concatenated to both streams (ablation) | MOCA · ALFRED val seen | task success rate: 25.85 -> 26.34 | No clear effect |
Closest papers by meaning
- ThinkBot: Embodied Instruction Following with Thought Chain Reasoning (ICLR 2025 · similarity 0.87)
- Generalizing Verifiable Instruction Following (NeurIPS 2025 · similarity 0.86)
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.85)
- Implicit Intelligence - Evaluating Agents on What Users Don’t Say (ICML 2026 · similarity 0.85)
- RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users (AAAI 2026 · similarity 0.85)
- Beyond Single Stationary Policies: Meta-Task Players as Naturally Superior Collaborators (NeurIPS 2024 · similarity 0.85)
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.
