Search papers, labs, and topics across Lattice.
CausalDS is a novel benchmark designed to evaluate causal reasoning within data-science workflows of large language models (LLMs) by integrating structural causal models with generated observational data and synthetic narratives. This benchmark addresses the limitations of existing causal evaluation datasets by systematically generating diverse synthetic causal structures rather than relying on curated examples, thereby reducing the risk of models merely mimicking existing data. Key findings indicate that CausalDS effectively assesses multiple dimensions of reasoning, including symbolic causal reasoning, uncertainty quantification, and the ability to abstain from unwarranted answers, enhancing the evaluation of LLMs in complex data-science tasks.
CausalDS reveals that LLMs can navigate complex causal reasoning tasks while effectively managing uncertainty and abstention, a critical skill for real-world data analysis.
Large language models (LLMs) increasingly act as integrated data-science agents, combining abstract reasoning with advanced tool use. Yet the relevant benchmark landscape largely divides into symbolic causal reasoning benchmarks without realistic data analysis or data analysis benchmarks without a principled causal data-generating structure. Furthermore, existing causal evaluation datasets are often restricted to curated examples from existing sources, with diversity coming from limited templatized variations rather than from systematic generation of novel synthetic causal structures. We introduce CausalDS, a benchmark for evaluating causal reasoning in agentic data-science workflows. Each benchmark instance is a scene consisting of a sampled structural causal model (SCM) with generated observational data and an accompanying synthetic natural-language story grounded in a realistic domain. We optionally ground the composition of the benchmark components in empirical distributions obtained from real-world datasets, thus retaining empirical structure while reducing the"causal parrot"risk through completely synthetic generation. From each scene, we then derive tasks spanning all three of Pearl's rungs, with typical data-science prediction tasks appearing as Rung 1. Most tasks include a data science coding component, where the model typically needs to use several tools to arrive at the final answer due to the frequent presence of imperfect observations, which are generated by an observation model. Additionally, recognizing when a question admits no warranted answer and abstaining is treated as a first-class scored outcome. The benchmark thus jointly evaluates symbolic causal reasoning, data science, uncertainty quantification, abstention, and tool use/coding.
