When Prompts Interact: Assessing Prompt Arithmetic for Deconfounding under Distribution Shift

arXiv:2605.03096 · cs.LG, cs.CL · Submitted 2026-05-04 · Read on arXiv

cs.LG, cs.CL

Submitted: 2026-05-04

Updated: 2026-09-01

Comments: 20 pages, 11 figures, 4 tables

Code: https://github.com/propublica/compas-analysis

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Task Arithmetic — 31× in this paper
• Robustness — 30× in this paper · explained in 5 episodes
• Fine-Tuning — 20× in this paper · explained in 5 episodes
• Prompt Arithmetic — 18× in this paper
• Distribution Shift — 16× in this paper · explained in 5 episodes
• Adjusted Integral — 11× in this paper
• Benchmark — 10× in this paper · explained in 2 episodes
• Embeddings — 10× in this paper · explained in 2 episodes
• Accuracy — 8× in this paper · explained in 1 episode
• Ablation — 6× in this paper · explained in 1 episode
• Generalization — 5× in this paper · explained in 13 episodes
• Adaptation — 3× in this paper · explained in 1 episode
• Neural Tangent Kernel — 3× in this paper
• Area Under the Shift Curve — 2× in this paper
• Calibration — 2× in this paper · explained in 7 episodes
• Covariate Shift — 2× in this paper · explained in 2 episodes
• Loss Function — 2× in this paper · explained in 2 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• AUC — 1× in this paper · explained in 2 episodes
• Ablation Studies — 1× in this paper · explained in 1 episode
• Active Learning — 1× in this paper · explained in 6 episodes
• Agentic Coding — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Area Under the Precision and Recall Curve — 1× in this paper
• Core — 1× in this paper · explained in 2 episodes
• Downstream Tasks — 1× in this paper · explained in 2 episodes
• Generalizability — 1× in this paper · explained in 6 episodes
• Latent Space — 1× in this paper · explained in 7 episodes
• Linear probe — 1× in this paper · explained in 2 episodes
• Modularity — 1× in this paper · explained in 9 episodes
• Provenance — 1× in this paper · explained in 4 episodes
• Regularization — 1× in this paper · explained in 3 episodes
• Social History Annotation Corpus — 1× in this paper
• Statistical Parity — 1× in this paper
• Transformer Architecture — 1× in this paper · explained in 2 episodes

Sources

• Invariant Risk Minimization
• Hate Speech Dataset from a White Supremacy Forum
• In Search of Lost Domain Generalization
• Editing Models with Task Arithmetic
• Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task Arithmetic
• Optimal Representations for Covariate Shift
• Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization
• Qwen3 Technical Report
• Change is Hard: A Closer Look at Subpopulation Shift

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks