CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference
CausalForge uses Lean proof checking to police automated causal-inference research, then adds a separate audit for whether the formal theorem matches the informal claim.
The framework pairs Causalean, a Lean library with 7,035 machine-checked declarations, with CausalSmith, an agentic pipeline for choosing topics, proposing results, formalizing them, proving them, and preparing artifacts for review. The authors argue that LLM-only reviewers are too unreliable for closing the research loop. They evaluate the system on artifacts from completed autonomous research runs. ArXiv · AI/CL/LG's note
The framework pairs Causalean, a Lean library with 7,035 machine-checked declarations, with CausalSmith, an agentic pipeline for choosing topics, proposing results, formalizing them, proving them, and preparing artifacts for review. The authors argue that LLM-only reviewers are too unreliable for closing the research loop. They evaluate the system on artifacts from completed autonomous research runs. ArXiv · AI/CL/LG's note
score 6