Submitted to ICASSP 2027 ยท submission
What Language Model Agents Need to Reason Causally over Multivariate Time-Series Signals
Benchmark and edge-ledger protocol for LLM causal-discovery agents on multivariate time series against PCMCI comparators: what the agents get right, and where they fail.
My part. Agentic causal discovery (PCMCI/Tigramite, edge-ledger).
Benchmark and edge-ledger protocol for LLM causal-discovery agents on multivariate time series against PCMCI comparators: what the agents get right, and where they fail.
Under review. Listed as a submission; nothing here claims acceptance.