Research map / Probabilistic and scientific modeling / Causal discovery
Pairwise Causality Guided Transformers for Event Sequences
Xiao Shou, Debarun Bhattacharjya, Tian Gao, Dharmashankar Subramanian, Oktie Hassanzadeh, Kristin Bennett
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 one pairwise causal statement via PC-TES loss | TES transformer · Synth-1 | test log-likelihood: -107.81 -> -107.19 | Improved |
| add one pairwise causal statement via PC-TES loss | TES transformer · Synth-2 | test log-likelihood: -114.01 -> -111.89 | Improved |
| add one pairwise causal statement via PC-TES loss | TES transformer · Synth-3 | test log-likelihood: -119.99 -> -113.58 | Improved |
| add one pairwise causal statement via PC-TES loss | TES transformer · Synth-4 | test log-likelihood: -113.95 -> -113.15 | No clear effect |
| add causal statements via PC-TES | TES transformer · Defi | test log-likelihood: -126.56 -> -120.54 | Improved |
| add causal statements via PC-TES | TES transformer · Diabetes | test log-likelihood: -373.27 -> -370.64 | Improved |
| add causal statements via PC-TES | TES transformer · Linkedin | test log-likelihood: -22.48 -> -21.71 | Improved |
| add causal statements via PC-TES | TES transformer · Beigebooks | test log-likelihood: -10.27 -> -10.04 | Improved |
| use transformer TES instead of attention-based PAIN | event sequence model · Synth-1 | test log-likelihood: -141.06 -> -107.81 | Improved |
| use neural models instead of parametric BSuMM/OSuMM/kMC | event sequence model · Synth-1 | test log-likelihood: -373.85 -> -107.81 | Improved |
| inject tsunami -> earthquake pair into PC-TES | PC-TES · LLM-generated event sequences | test log-likelihood: -28.29 -> -27.48 | Improved |
| inject impeachment -> protest pair into PC-TES | PC-TES · LLM-generated event sequences | test log-likelihood: -28.29 -> -28.19 | No clear effect |
Closest papers by meaning
- Towards Understanding the Universality of Transformers for Next-Token Prediction (ICLR 2025 · similarity 0.86)
- MOCHA: Discovering Multi-Order Dynamic Causal Structure in Temporal Point Processes (NeurIPS 2026 · similarity 0.86)
- Linear Attention for Efficient Bidirectional Sequence Modeling (NeurIPS 2025 · similarity 0.85)
- On the Identification of Temporal Causal Representation with Instantaneous Dependence (ICLR 2025 · similarity 0.85)
- UnCLe: Towards Scalable Dynamic Causal Discovery in Non-linear Temporal Systems (NeurIPS 2025 · similarity 0.85)
- Causal Order: The Key to Leveraging Imperfect Experts in Causal Inference (ICLR 2025 · similarity 0.85)
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.
