Research map / Probabilistic and scientific modeling / Treatment effects
Learning Treatment Representations for Downstream Instrumental Variable Regression
Shiangyi Lin, Hui Lan, Vasilis Syrgkanis
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 instrument-regression term mu1 to the lambda-only regularizer (IRAE[0] to IRAE[1]) (ablation) | IRAE · Quadratic data, Case 2 | Average test improvement: 2.2898 -> 4.6175 | Improved |
| add mu1 term (IRAE[0] to IRAE[1]) (ablation) | IRAE · Quadratic data, Case 1 | Average test improvement: 6.1055 -> 6.4174 | No clear effect |
| add mu2 and larger bottleneck 10 (IRAE[1] to IRAE[2]) (ablation) | IRAE · Quadratic data, Case 2 | Average test improvement: 4.6175 -> 4.5554 | No clear effect |
| add mu3 D-V dependence penalty (IRAE[2] to IRAE) (ablation) | IRAE · Quadratic data, Case 2 | Average test improvement: 4.5554 -> 6.5253 | Improved |
| add mu3 D-V dependence penalty (IRAE[2] to IRAE) (ablation) | IRAE · Quadratic data, Case 3 | Average test improvement: 5.2145 -> 4.9113 | Hurt |
| add mu3 D-V dependence penalty (IRAE[2] to IRAE) (ablation) | IRAE · MNIST (sample size 10000, intervened alpha=1.0) | Average test improvement: 0.92 -> 1.1 | Improved |
| add instrument regularization with bottleneck 10 (IRAE[1] to IRAE[2]) (ablation) | IRAE · MNIST (sample size 10000, intervened alpha=1.0) | Average test improvement: 0.15 -> 0.92 | Improved |
| use LIRR instead of PCA | LIRR · Linear data, m=50, DGP 3 | Average test improvement: 2.5171 -> 5.4944 | Improved |
| use LIRR instead of PCA with independent Gaussian noise | LIRR · Linear data, m=500, DGP 1 | Average test improvement: 1.0098 -> 1.0355 | No clear effect |
| increase dimension of X from 50 to 500 | LIRR · Linear data, DGP 1 | Average test improvement: 3.7283 -> 1.0355 | Hurt |
| use IRAE instead of vanilla AE | IRAE · Quadratic data, Case 1 | Average test improvement: 0.4138 -> 5.7740 | Improved |
| use LIRR instead of iVAE | LIRR · Quadratic data, Case 2 | Average test improvement: 0.1192 -> 3.4711 | Improved |
Closest papers by meaning
- Learning Representations of Instruments for Partial Identification of Treatment Effects (ICML 2025 · similarity 0.86)
- Conditional Instrumental Variable Regression with Representation Learning for Causal Inference (ICLR 2024 · similarity 0.86)
- Demystifying Spectral Feature Learning for Instrumental Variable Regression (NeurIPS 2025 · similarity 0.86)
- Meta-Learners for Partially-Identified Treatment Effects Across Multiple Environments (ICML 2024 · similarity 0.86)
- An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation (NeurIPS 2025 · similarity 0.85)
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data (NeurIPS 2025 · similarity 0.84)
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.
