Research map / Probabilistic and scientific modeling
Causal discovery: research map
973 papers on Causal discovery in Probabilistic and scientific modeling, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 10 approaches. The busiest year so far is 2026.
What it is
Causal discovery tries to recover which variables cause which from data, usually as a directed graph. It is hard because correlation does not settle direction and hidden variables can confound; methods use statistical tests, interventions or score-based search.
Terms you will see on this page:
- DAG
- directed acyclic graph, the usual picture of causal relations.
- Confounder
- a hidden cause of two variables that makes them look related.
- Identifiability
- whether the true causal structure can be recovered from the available data at all.
Start reading with the most cited papers here:
- CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models (CVPR 2021 · 248 citations)
- Discovering Causal Signals in Images (CVPR 2017 · 206 citations)
Within Probabilistic and scientific modeling, its share held steady from 11.4% in 2023–24 to 11.0% in 2025–26 (232 → 334 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Causal discovery in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Counterfactual identification and bounds rising: 7% → 14% of the topic (16 → 46 papers)
- Causal discovery from time series rising: 6% → 7% of the topic (13 → 25 papers)
- Identifiability with hidden confounders rising: 9% → 11% of the topic (21 → 36 papers)
- Causal effect identification falling: 16% → 8% of the topic (37 → 27 papers)
- Causal representation identifiability falling: 17% → 16% of the topic (40 → 52 papers)
Papers from 2025–2026 that are already cited most:
- Image Quality Assessment: Investigating Causal Perceptual Effects with Abductive Counterfactual Inference (CVPR 2025 · 25 citations)
- A Practical Approach to Causal Inference over Time (AAAI 2025 · 3 citations)
- Sequential Conditional Transport on Probabilistic Graphs for Interpretable Counterfactual Fairness (AAAI 2025 · 1 citations)
Approaches and key papers
Counterfactuals and causal representations · 547 papers
Keywords: counterfactual, causal inference, causal representation
Approaches in this cluster:
- Latent-variable causal estimation (132 papers)
Learn structural causal models and deconfounded estimates using generative and latent-variable models. - Causal effect identification (129 papers)
Derive identifiability conditions and estimators for causal effects under latent confounding. - Causal representation identifiability (141 papers)
Prove when causal variables can be recovered from observations under interventions or weak supervision. - Counterfactual identification and bounds (95 papers)
Characterize and bound counterfactual quantities from observational and experimental data. - Causal approaches to fairness (50 papers)
Apply causal graphs and interventional criteria to define and enforce fairness.
Most cited and most cited since 2024:
- CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models (CVPR 2021 · 248 citations)
- Learning Representations for Counterfactual Inference (ICML 2016 · 185 citations)
- Robust Emotion Recognition in Context Debiasing (CVPR 2024 · 29 citations)
- Image Quality Assessment: Investigating Causal Perceptual Effects with Abductive Counterfactual Inference (CVPR 2025 · 25 citations)
Learning causal graphs · 426 papers
Keywords: causal discovery, causal structure, graphs
Approaches in this cluster:
- Constraint-based structure learning with latents (108 papers)
Recover causal graphs under latent variables and mixed data using recursive, constraint-based algorithms. - Identifiability with hidden confounders (100 papers)
Prove and exploit conditions for discovering causal structure under latent confounding and measurement error. - Interventional causal discovery (83 papers)
Use interventional data and targeting strategies to learn causal graphs efficiently. - Differentiable and score-based DAG learning (77 papers)
Learn causal DAGs by continuous optimization, reinforcement learning, or noise-model assumptions. - Causal discovery from time series (58 papers)
Recover causal structure from temporal data under non-stationarity, latent confounding and Granger-style assumptions.
Most cited and most cited since 2024:
- Discovering Causal Signals in Images (CVPR 2017 · 206 citations)
- Permutation-based Causal Inference Algorithms with Interventions (NeurIPS 2017 · 56 citations)
- Causal-learn: Causal Discovery in Python (NeurIPS 2024 · 26 citations)
- Identification of Causal Structure in the Presence of Missing Data with Additive Noise Model (AAAI 2024 · 8 citations)
Related topics in Probabilistic and scientific modeling
- Dynamical systems (1,019)
- Uncertainty and Gaussian processes (1,421)
- Molecules and proteins (2,120)
- Variational and generative sampling (2,370)
- Treatment effects (468)
- Differential equations and PDEs (676)
- Bayesian optimization (434)
