Research map / Large language models / Retrieval and knowledge
A Dataset for Answering Time-Sensitive Questions
Wenhu Chen, Xinyi Wang, William Yang Wang
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 |
|---|---|---|---|
| Continue fine-tuning BigBird (NQ) on TimeQA | BigBird (FT on NQ) · TimeQA Test-Easy | EM: 28.6 -> 51.2 | Improved |
| Continue fine-tuning BigBird (NQ) on TimeQA | BigBird (FT on NQ) · TimeQA Test-Hard | EM: 25.5 -> 42.4 | Improved |
| Continue fine-tuning BigBird (TriviaQA) on TimeQA | BigBird (FT on TriviaQA) · TimeQA Test-Easy | EM: 33.7 -> 50.8 | Improved |
| Continue fine-tuning BigBird (TriviaQA) on TimeQA | BigBird (FT on TriviaQA) · TimeQA Test-Hard | EM: 27.7 -> 40.4 | Improved |
| Continue fine-tuning FiD (NQ) on TimeQA | FiD (FT on NQ) · TimeQA Test-Easy | EM: 23.3 -> 60.5 | Improved |
| Continue fine-tuning FiD (NQ) on TimeQA | FiD (FT on NQ) · TimeQA Test-Hard | EM: 16.0 -> 46.8 | Improved |
| Continue fine-tuning FiD (TriviaQA) on TimeQA | FiD (FT on TriviaQA) · TimeQA Test-Easy | EM: 23.2 -> 57.5 | Improved |
| Continue fine-tuning FiD (TriviaQA) on TimeQA | FiD (FT on TriviaQA) · TimeQA Test-Hard | EM: 13.1 -> 46.2 | Improved |
| Evaluate on hard rather than easy questions | FiD (FT on NQ + TimeQA) · TimeQA test split | EM: 60.5 -> 46.8 | Hurt |
| Add human-paraphrased TimeQA training examples | BigBird (FT on NQ + TimeQA) · Human-paraphrased close-domain Test-Easy | EM: 45.5 -> 47.6 | Improved |
| Add human-paraphrased TimeQA training examples | BigBird (FT on NQ + TimeQA) · Human-paraphrased close-domain Test-Hard | EM: 36.5 -> 38.8 | Improved |
Closest papers by meaning
- TimE: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-World Scenarios (NeurIPS 2025 · similarity 0.88)
- ITFormer: Bridging Time Series and Natural Language for Multi-Modal QA with Large-Scale Multitask Dataset (ICML 2025 · similarity 0.87)
- Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning (ICLR 2025 · similarity 0.87)
- ListQA: A Benchmark for Evaluating List-Formatted Factual Knowledge Retrieval in Large Language Models (NeurIPS 2026 · similarity 0.84)
- ReXTime: A Benchmark Suite for Reasoning-Across-Time in Videos (NeurIPS 2024 · similarity 0.83)
- Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · similarity 0.83)
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.
