@adiba_ejaz: How can causal (and statistical) models generalize to novel combinations of interacting objects? Our work w/ @eliasbare…
Summary
This paper presented at ICML explores how causal and statistical models can generalize to novel combinations of interacting objects, with a poster session scheduled at the conference.
View Cached Full Text
Cached at: 07/04/26, 08:53 PM
How can causal (and statistical) models generalize to novel combinations of interacting objects? Our work w/ @eliasbareinboim at @icmlconf considers this question. Happy to chat about the paper and causal world models more broadly!
Poster: Wed Jul 8 5-6:45 PM KST Hall A #4207 https://t.co/EO1xhZaW4m
Similar Articles
@eliasbareinboim: Looking forward to reconnecting with many of you and meeting new colleagues at #ICML in Seoul. Across several of our pa…
Elias Bareinboim announces his group's multiple ICML 2025 papers on causal AI, covering relational world models, robust offline RL, counterfactual identification, and causal game theory, highlighting the need for causal knowledge in AI reasoning.
CausaLab: A Scalable Environment for Interactive Causal Discovery Toward AI Scientists
CausaLab is a scalable environment for evaluating LLM agents on interactive causal discovery, assessing both predictive accuracy and faithful recovery of underlying causal mechanisms. Experiments reveal a gap between prediction and mechanism recovery, highlighting limits in current LLM agents as experimental causal reasoners.
Relational Structural Causal Models
This paper introduces relational structural causal models, extending structural causal models to settings with varying objects and relations. It provides theoretical results for identification and proposes relational neural causal models that outperform non-relational baselines on simulated traffic scenes.
Interpretable Causal Discovery via Causal-Effect Constraints
This paper introduces a Bayesian approach to conditional causal discovery, where the posterior over causal graphs and parameters is conditioned on user-specified causal-effect constraints (e.g., a large causal effect). They adapt rare-event estimation techniques to handle events with small posterior mass and validate the method on synthetic data and the Sachs protein dataset.
CausalDS: Benchmarking Causal Reasoning in Data-Science Agents
Introduces CausalDS, a benchmark for evaluating causal reasoning in LLM-based data science agents, using synthetic structural causal models and natural language stories to test associational, interventional, and counterfactual reasoning along with tool use and abstention.