Research map / Large language models / Retrieval and knowledge
StepGame: A New Benchmark for Robust Multi-Hop Spatial Reasoning in Texts
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 TP-MANN instead of the strongest reported baseline | bAbI task 17 · bAbI task 17 | accuracy: 98.60$\pm$3.40 -> 99.88$\pm$0.10 | Helped |
| Use TP-MANN instead of STM | bAbI task 19 · bAbI task 19 | accuracy: 99.98$\pm$0.05 -> 99.98$\pm$0.04 | No clear effect |
| Use TP-MANN instead of TPR-RNN | StepGame · StepGame k=1–5, noisy test | mean accuracy: 40.81 -> 52.99 | Helped |
| Use TP-MANN instead of TPR-RNN | StepGame · StepGame k=6–10, noisy test | mean accuracy: 16.65 -> 24.53 | Helped |
| Increase the number of recurrent memory layers (ablation) | TP-MANN · StepGame, especially larger k | accuracy | Helped |
| Increase the number of recurrent memory layers (ablation) | TP-MANN · StepGame k=6–10, noisy test | accuracy | Helped |
Closest papers by meaning
- Advancing Spatial Reasoning in Large Language Models: An In-Depth Evaluation and Enhancement Using the StepGame Benchmark (AAAI 2024 · similarity 0.92)
- ChainSpace: A Chained-Reasoning Paradigm for Spatial Intelligence (NeurIPS 2026 · similarity 0.88)
- Enhancing Spatial Reasoning Through Visual and Textual Thinking (AAAI 2026 · similarity 0.87)
- SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence (CVPR 2026 · similarity 0.87)
- SpatiaLQA: A Benchmark for Evaluating Spatial Logical Reasoning in Vision-Language Models (CVPR 2026 · similarity 0.87)
- Spatial-DISE: A Unified Benchmark for Evaluating Spatial Reasoning in Vision-Language Models (ICLR 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.