Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Housekeep: Tidying Virtual Households Using Commonsense Reasoning
Yash Kant, Arun Ramachandran, Sriram Yenamandra, Igor Gilitschenski, Dhruv Batra, Andrew Szot, Harsh Agrawal
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 |
|---|---|---|---|
| fine-tune embeddings with contrastive matching (CM) instead of zero-shot MLM | RoBERTa ranker · Housekeep ORR test-unseen | mAP: 0.42 -> 0.81 | Improved |
| fine-tune embeddings with contrastive matching (CM) instead of zero-shot MLM | RoBERTa ranker · Housekeep OR test-unseen | mAP: 0.52 -> 0.65 | Improved |
| use RoBERTa instead of GloVe embeddings with CM | CM ranker · Housekeep ORR test-unseen | mAP: 0.76 -> 0.81 | Improved |
| use RoBERTa instead of GloVe embeddings with CM | CM ranker · Housekeep OR test-unseen | mAP: 0.66 -> 0.65 | No clear effect |
| zero-shot MLM ranking instead of random ranking | LM ranker · Housekeep OR test-unseen | mAP: 0.59 -> 0.52 | Hurt |
| replace oracle ranker with LLM ranker | Oracle exploration agent · Housekeep test-seen | Object Success (OS): 1.00 -> 0.44 | Hurt |
| replace oracle ranker with LLM ranker | Oracle exploration agent · Housekeep test-seen | Episode Success (ES): 1.00 -> 0.04 | Hurt |
| replace oracle exploration with frontier exploration | Oracle ranker agent · Housekeep test-seen | Object Success (OS): 1.00 -> 0.64 | Hurt |
| replace oracle exploration with frontier exploration | Oracle ranker agent · Housekeep test-seen | Episode Success (ES): 1.00 -> 0.35 | Hurt |
| replace oracle exploration with frontier exploration | LM ranker agent · Housekeep test-seen | Object Success (OS): 0.44 -> 0.30 | Hurt |
| evaluate on unseen instead of seen object categories | Full LM + frontier agent · Housekeep | Object Success (OS): 0.30 -> 0.23 | Hurt |
| evaluate on unseen instead of seen object categories | Full LM + frontier agent · Housekeep | Soft Object Success (SOS): 0.39 -> 0.36 | No clear effect |
Closest papers by meaning
- TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning (NeurIPS 2026 · similarity 0.89)
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.88)
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (ICML 2025 · similarity 0.87)
- Tokenization Allows Multimodal Large Language Models to Understand, Generate and Edit Architectural Floor Plans (CVPR 2026 · similarity 0.86)
- Embodied Web Agents: Bridging Physical-Digital Realms for Integrated Agent Intelligence (NeurIPS 2025 · similarity 0.86)
- LoTa-Bench: Benchmarking Language-oriented Task Planners for Embodied Agents (ICLR 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.
