Research map / Probabilistic and scientific modeling / Causal discovery
Controllable Sequence Editing for Biological and Clinical Trajectories
Michelle M Li, Kevin Li, Yasha Ektefaie, Ying Jin, Yepeng Huang, Shvat Messica, Tianxi Cai, Marinka Zitnik
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 CLEF temporal concept encoder/decoder to sequence encoders | Transformer, xLSTM, MOMENT · Immediate sequence editing, 8 datasets | MAE: 16.28% lower MAE on average | Improved |
| add CLEF temporal concepts | Transformer, xLSTM, MOMENT · Delayed sequence editing | MAE: 26.73% lower MAE on average | Improved |
| add CLEF temporal concepts | sequence encoders · Zero-shot counterfactual cellular trajectories (WOT-CF) | MAE: up to 62.84% lower MAE | Improved |
| add CLEF to MOMENT encoder | MOMENT · eICU, MIMIC-IV, WOT | MAE | Improved |
| use CLEF instead of non-CLEF | Transformer, xLSTM · eICU, MIMIC-IV challenging splits | generalization to new patients | Improved |
| learned temporal concepts instead of all-ones concepts (SimpleLinear) (ablation) | CLEF · WOT immediate and delayed editing | MAE | Mixed |
| add CLEF to CT/CRN with high time-varying confounding (gamma >= 3) | CT with CDC, CRN with GR · tumor growth, ICU counterfactual prediction | tau-step ahead error | Improved |
| add CLEF to CT/CRN with low time-varying confounding (gamma < 3) | CT with CDC, CRN with GR · tumor growth, ICU counterfactual prediction | tau-step ahead error | No clear effect |
| remove balancing loss (GR or CDC) from CLEF-CT/CRN | CLEF-CT, CLEF-CRN · tumor growth, ICU counterfactual prediction | tau-step ahead error | Improved |
| add three intermediate autoregressive steps before predicting t_j (ablation) | CLEF · eICU, MIMIC-IV delayed editing | MAE | Hurt |
Closest papers by meaning
- Conditional Recurrent Flow: Conditional Generation of Longitudinal Samples With Applications to Neuroimaging (ICCV 2019 · similarity 0.84)
- Finding Counterfactually Optimal Action Sequences in Continuous State Spaces (NeurIPS 2023 · similarity 0.83)
- ConTSG-Bench: A Unified Benchmark for Conditional Time Series Generation (ICML 2026 · similarity 0.83)
- STRIDE: Post-Training LLMs to Reason and Refine Bio-Sequences via Edit Trajectories (ICML 2026 · similarity 0.83)
- Forward-Chaining Temporal Point Process (ICML 2026 · similarity 0.83)
- Variational Counterfactual Intervention Planning to Achieve Target Outcomes (ICML 2025 · similarity 0.82)
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.
