1 repository on SrcLog
Add it to a GitHub README or profile β the card updates as the ranking changes:
Counterfactual replay and causal attribution over LLM-agent trajectories: find which step actually caused a failure by intervening on it.