Research map / Large language models / Agents, code and math
PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change
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 |
|---|---|---|---|
| obfuscate domain names with misleading words (ablation) | GPT-4 · Blocksworld plan generation | instances correct (%): 34.3 -> 4.3 | Hurt |
| obfuscate domain names with random strings (ablation) | GPT-4 · Blocksworld plan generation | instances correct (%): 34.3 -> 2 | Hurt |
| obfuscate domain names with misleading words (ablation) | InstructGPT-3 · Blocksworld plan generation | instances correct (%): 6.8 -> 0.23 | Hurt |
| obfuscate domain names with random strings (ablation) | InstructGPT-3 · Blocksworld plan generation | instances correct (%): 6.8 -> 1 | Hurt |
| use GPT-4 instead of InstructGPT-3 | LLM · Blocksworld plan generation | instances correct (%): 6.8 -> 34.3 | Helped |
| use GPT-4 instead of InstructGPT-3 | LLM · Blocksworld plan verification | instances correct (%): 12 -> 58.6 | Helped |
| use GPT-4 instead of InstructGPT-3 | LLM · Goal reformulation (shuffle goals) | instances correct (%): 77.8 vs 76.8 | No clear effect |
| use GPT-4 instead of InstructGPT-3 | LLM · Goal reformulation (partial to full) | instances correct (%): 60.5 vs 58 | No clear effect |
| use GPT-4 instead of InstructGPT-3 | LLM · Goal reformulation (full to partial) | instances correct (%): 77.8 -> 87 | Helped |
Closest papers by meaning
- ACPBench: Reasoning About Action, Change, and Planning (AAAI 2025 · similarity 0.93)
- LLMs Can Plan Only If We Tell Them (ICLR 2025 · similarity 0.92)
- TravelPlanner: A Benchmark for Real-World Planning with Language Agents (ICML 2024 · similarity 0.91)
- Large Language Models as Planning Domain Generators (Student Abstract) (AAAI 2024 · similarity 0.91)
- Graph-enhanced Large Language Models in Asynchronous Plan Reasoning (ICML 2024 · similarity 0.91)
- Planning in the Dark: LLM-Symbolic Planning Pipeline Without Experts (AAAI 2025 · similarity 0.91)
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.