Research map / Vision-language and multimodal / Robot action and navigation
CoReLIN: Constraint-based Reasoning for Zero-shot Lifelong Interactive Navigation
Apoorva Vashisth, Manav Kulshrestha, Pranav Bakshi, Damon Conover, Guillaume Sartoretti, Aniket Bera
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 CoReLIN (known) instead of ADIN | ADIN · ProcTHOR-10k, g=20, 7-10 rooms | LES: 12.21 -> 86.56 | Improved |
| use CoReLIN (known) instead of Always Interact in large scenes | Always Interact · ProcTHOR-10k, g=20, 7-10 rooms | LES: 74.73 -> 86.56 | Improved |
| use CoReLIN (known) instead of Clean + S/P in large scenes | Clean + S/P · ProcTHOR-10k, g=20, 7-10 rooms | LES: 61.86 -> 86.56 | Improved |
| use CoReLIN (known) instead of Always Interact in small scenes | Always Interact · ProcTHOR-10k, g=20, 1-3 rooms | LES: 94.18 -> 89.20 | Hurt |
| remove ground-truth knowledge (unk vs known) (ablation) | CoReLIN · ProcTHOR-10k, g=20, 7-10 rooms | LES: 86.56 -> 49.99 | Hurt |
| use CoReLIN (unk) instead of InterNav | InterNav · ProcTHOR-10k, g=20, 7-10 rooms | LES: 0.20 -> 49.99 | Improved |
| increase manipulation cost e from 1 to 20 (ablation) | CoReLIN (unk) · ProcTHOR-10k, 1-3 rooms | interaction encounters (IE): 42.21 -> 9.58 | Improved |
| increase manipulation cost e from 1 to 20 (ablation) | CoReLIN (unk) · ProcTHOR-10k, 7-10 rooms | PoC: 2.65 -> 2.95 | Hurt |
| increase manipulation cost e from 1 to 20 (ablation) | CoReLIN (unk) · ProcTHOR-10k, 1-3 rooms | SR (%): 92.58 -> 92.88 | No clear effect |
| increase history length h from 1 to 9 (ablation) | CoReLIN (unk) · ProcTHOR-10k, 4-6 rooms | LES: 71.84 -> 75.82 | Improved |
| increase history length h from 1 to 3 (ablation) | CoReLIN (unk) · ProcTHOR-10k, 7-10 rooms | LES: 41.37 -> 52.60 | Improved |
| use Gemini instead of GPT-5 as LLM (ablation) | CoReLIN (unk) · ProcTHOR-10k, 4-6 rooms | LES: 53.11 -> 68.44 | Improved |
Closest papers by meaning
- VoroNav: Voronoi-based Zero-shot Object Navigation with Large Language Model (ICML 2024 · similarity 0.86)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.86)
- STRIDER: Navigation via Instruction-Aligned Structural Decision Space Optimization (NeurIPS 2025 · similarity 0.85)
- C-NAV: Towards Self-Evolving Continual Object Navigation in Open World (NeurIPS 2025 · similarity 0.85)
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning (NeurIPS 2026 · similarity 0.84)
- GOAT-Bench: A Benchmark for Multi-Modal Lifelong Navigation (CVPR 2024 · similarity 0.84)
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.
