Research map / Time series / Causal and generative time series
What if Tomorrow is the World Cup Final? Counterfactual Time Series Forecasting with Textual Conditions
Shuqi Gu, Yongxiang Zhao, Baoyu Jing, Kan Ren
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 |
|---|---|---|---|
| add text-attribution (use historical intrinsic features as initial noise) (ablation) | TADiff · Traffic factual | MSE: 0.78 -> 0.27 | Improved |
| add text-attribution (ablation) | TADiff · ETTm1 counterfactual | DTTC-I: 9.35 -> 10.59 | Improved |
| add text-attribution (ablation) | TADiff · Traffic counterfactual | DTTC-E: 73.52 -> 78.58 | Improved |
| add text-attribution (ablation) | TADiff · ETTm1 factual | MSE: 0.77 -> 0.76 | No clear effect |
| add counterfactual finetuning (ablation) | TADiff · ETTm1 counterfactual | DTTC-I: 10.06 -> 10.59 | Improved |
| add counterfactual finetuning (ablation) | TADiff · ETTm1 factual | MSE: 0.82 -> 0.76 | Improved |
| add counterfactual finetuning (ablation) | TADiff · Traffic factual | MSE: 0.24 -> 0.27 | Hurt |
| add counterfactual finetuning (ablation) | TADiff · Traffic counterfactual | DTTC-I: 17.31 -> 18.08 | Mixed |
| use TADiff instead of VerbalTS generation baseline | text-conditioned forecasting · ETTm1 factual | MSE: 0.87 -> 0.76 | Improved |
| use TADiff instead of IATSF | Weather factual · Weather | MSE: 0.09 -> 0.11 | Hurt |
| vary loss weight ratio lambda_F:lambda_A and lambda_I:lambda_E (ablation) | TADiff training · Synth / ETTm1 | MAE, MSE, DTTC | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VerbalTS — VerbalTS: Generating Time Series from Texts (ICML 2025) · better on 5 of 5 shared results
- Sundial — Sundial: A Family of Highly Capable Time Series Foundation Models (ICML 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Context is Key: A Benchmark for Forecasting with Essential Textual Information (ICML 2025 · similarity 0.84)
- Online Time Series Forecasting with Theoretical Guarantees (NeurIPS 2025 · similarity 0.83)
- Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis (NeurIPS 2024 · similarity 0.83)
- ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data (AAAI 2025 · similarity 0.83)
- Introducing FOReCAst: The Future Outcome Reasoning and Confidence Assessment Benchmark (NeurIPS 2025 · similarity 0.83)
- AutoTimes: Autoregressive Time Series Forecasters via Large Language Models (NeurIPS 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.
