Tag
Introduces Tree-Coupled A/B Testing (TCAB), an exact feedback-sharing design for comparing multiple adaptive policies with fewer reward queries while preserving each policy's trajectory law.
A controlled, reproducible benchmark evaluating offline estimators for equal-cost top-k allocation, identifying when offline evaluation can be trusted and providing a practitioner's guide.
This paper derives tighter bounds for multivalued probabilities of causation by incorporating causal information from covariates and mediators, extending prior work from binary settings.
Presents FunnelCausalNet, an uplift estimation method for multi-tier coupon allocation that couples conversion and revenue heads under a funnel composition to reduce variance, validated on semi-synthetic Criteo and industrial hotel-coupon RCT logs.
This paper introduces the Counterfactual Clinical Audit (CCA) framework to evaluate offline reinforcement learning agents for ICU sepsis management, exposing 'toxic mimicry' where agents replicate harmful treatment patterns that standard metrics miss. Using MIMIC-III data, it shows a Medical Decision Transformer fails to escalate vasopressors under rising lactate, while a causal transformer performs safely.
This paper presents a method using Group Relative Policy Optimization (GRPO) to fine-tune an open-weight language model for generating actionable financial advice, outperforming commercial LLMs under a judge-independent CATE evaluation while also matching safety criteria.
This paper introduces CAR-PL, an observational policy ranking method for recommending business changes to small and medium-sized businesses using multi-action accounting logs, comparing it against several baselines on financial KPI prediction.
This paper theoretically analyzes support selection in continuous DAG learning, showing that smooth acyclicity constraints alone cannot rank supports beyond feasibility and deriving selection times for NOTEARS/DAGMA, with empirical audits on 320 trajectories.
This paper introduces CLAM, a method for estimating localized causal effects from coarse-resolution data by jointly learning causal mechanisms and a disaggregation mapping, with applications in public health and environmental policy.
This paper introduces Target-Weighted Neyman Allocation (TWNA), a two-stage stratified experimental design that optimizes sample allocation across groups and treatment arms to improve precision of target-weighted group average treatment effects under population shift.
This paper investigates using GPT-3.5 for counterfactual analysis in online lending, showing that prompt engineering improves prediction accuracy and enables coherent counterfactual ROI generation under alternative interest rates.
This paper audits visual tool-use in multimodal LLMs via causal interventions, revealing that returned observations often lack causal effect despite aggregate accuracy gains. It identifies failure modes like 'Calling Without Looking' and 'Looking Without Planning', introducing the concept of the 'illusion of visual tool-use'.
UpliftBench is a benchmark paper showing that disagreements between uplift modeling evaluations often stem from metric choice rather than model quality, identifying specific mismatches between ranking metrics and deployment objectives across several dataset families.
DoTime is a synthetic benchmark generator for interventional and counterfactual time series, providing scalable TSCM-based data generation with exact ground truth, released as a PyPI package with evaluation suites. It enables training and benchmarking causal foundation models on non-observational time series data.
Presents Eco3S, a socio-economic system simulation framework that uses LLM-based agents with co-evolving environments, structural causal simulation, and a self-corrective refinement paradigm to replicate and analyze economic phenomena.
This paper proposes CURL, a plug-in adapter that uses estimator uncertainty to allocate pretrained LLM semantic capacity for improving heterogeneous treatment effect (CATE) estimation. It introduces two role-conditioned prompts to construct assignment- and heterogeneity-oriented representations, improving ten host learners on four benchmarks.
This paper evaluates unsupervised subgrouping methods combined with causal discovery and policy evaluation for budget-constrained health interventions using observational data, finding no single method consistently outperforms others in held-out evaluation.
This paper extends Pearl's structural causal model framework by introducing causal zeros and causal differential equations to handle symmetric constraints and feedback cycles, which are not allowed in directed acyclic graphs.
This paper presents a partial-identification approach to use prior RCT data to bound the causal effect of a new ML model, leveraging assumptions about counterfactual correctness and subgroup predictive accuracy to yield more informative bounds.
This paper studies contextual causal bandits with non-manipulable variables, proposing causal variants of Thompson Sampling and Information-Directed Sampling (IDS) that exploit shared causal mechanisms to accelerate decision-making. Theoretical regret bounds and experiments on synthetic tasks show that the proposed methods outperform causal and non-causal baselines.