CausalDS is a benchmark generator for causal reasoning in data-science agents. Each generated scene pairs a synthetic structural causal model with tabular data and an audited natural-language story. Tasks span association, intervention, and counterfactual reasoning, including estimation, uncertainty, causal structure, and knowing when to abstain. This release contains the 100-task realistic exam and the observation-layer and CauseNet-seeding ablations, including evaluation inputs, reference answers, and held-out test labels. The code repository provides the generator, agent harness, and deterministic grader. The CausalDSScore $\times$ $\log$ tokens per task plane separates the latest closed…
Independent publisher
Andrej Leban
andleb
Models in Library0
Datasets in Library1
Models on Hugging Face—
Followers—
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Other.
Reported licenses: cc0-1.0. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.