Research map / Large language models / Agents, code and math
SPAN: Benchmarking and Improving Cross-Calendar Temporal Reasoning of Large Language Models
Zhongjian Miao, Hao Fu, Chen Wei
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 |
|---|---|---|---|
| replace direct reasoning with tool-augmented code generation (Time Agent) | GPT-4o · SPAN | accuracy: 43.01 -> 95.31 | Improved |
| add Bing RAG retrieval (ablation) | GPT-4o · SPAN | accuracy: 43.01 -> 43.69 | No clear effect |
| use reasoning model o1 instead of GPT-4o | GPT-4o · SPAN | accuracy: 43.01 -> 59.29 | Improved |
| future vs past reference dates | six LLMs · SPAN | accuracy: e.g. GPT-4o 46.38 / 43.01 / 38.88 (annotated values) | Hurt |
| Gregorian-to-others vs others-to-Gregorian direction | six LLMs · SPAN | accuracy: gains of 3.97% to 17.49% for Gregorian source | Improved |
| polar instead of content question format | six LLMs · SPAN | accuracy: gains 3.13% to 37.08% (18.86% avg) | Improved |
| festival-based instead of date-based questions | six LLMs · SPAN | accuracy: gains 2.87% to 12.60% for most, not Gemini-1.5-Pro or Qwen | Mixed |
Closest papers by meaning
- TimE: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-World Scenarios (NeurIPS 2025 · similarity 0.92)
- Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning (ICLR 2025 · similarity 0.90)
- Multilingual LLMs Inherently Reward In-Language Time-Sensitive Semantic Alignment for Low-Resource Languages (AAAI 2025 · similarity 0.89)
- Period-LLM: Extending the Periodic Capability of Multimodal Large Language Model (CVPR 2025 · similarity 0.88)
- Plan and Budget: Effective and Efficient Test-Time Scaling on Reasoning Large Language Models (ICLR 2026 · similarity 0.88)
- NeSTR: A Neuro-Symbolic Abductive Framework for Temporal Reasoning in Large Language Models (AAAI 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.
