Andrej Leban
Andrej Leban
Home
Research
Publications
Talks
Honors & Media
Teaching
Posts
Projects
Light
Dark
Automatic
Synthetic Data
K2-Horizon: A Frontier Fleet of Open Source Language Models
Technical report for the K2-Horizon family of open-source language models. My main contribution is the
research lab
multi-agent simulation described in Sec. 5.12.5 of the report.
Zhengzhong Liu
,
…
,
Andrej Leban
,
et al.
Technical report
PDF
Cite
arXiv
alphaXiv
Hugging Face
CausalDS: Benchmarking Causal Reasoning in Data-Science Agents
CausalDS is a
benchmark generator
for agentic causal data science: every problem is generated fresh in its entirety, with tasks spanning all three rungs of Pearl’s hierarchy that involve significant tool use. Exam composition is a free parameter: it can be tailored to a specific goal or grounded in real-world corpora. The benchmark jointly tests symbolic causal reasoning, data-science execution, uncertainty quantification, epistemic abstention, and coding/tool use.
Andrej Leban
,
Theo Saarinen
,
Yuekai Sun
arXiv preprint
PDF
Cite
Code
arXiv
alphaXiv
Hugging Face dataset
New preprint: CausalDS
We have a new preprint out on benchmarking agentic causal data-science.
Last updated on Jul 9, 2026
1 min read
Cite
×