mlsanityData to 7 Oct 2026

Barriers to Counterfactual Credit Attribution for Autoregressive Models

arXiv 2605.01425 ↗First posted 2 May 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Generative AI disrupts the practice of giving credit to work that came before. Ideally, a generative model would give credit to any work on which its output depends in a significant way. Counterfactual credit attribution (CCA) is a technical condition formalizing this goal--a relaxation of differential privacy--recently introduced by Livni, Moran, Nissim, and Pabbaraju [2024] who studied it in the PAC learning setting. We initiate the study of CCA generative models. Specifically, we consider autoregressive models giving credit to a deployment-time dataset (e.g., a RAG database). We uncover barriers to two natural approaches to CCA autoregressive models. First, we show that imposing CCA on the underlying next-token predictor does not guarantee that the model is CCA: CCA does not compose autoregressively (unlike DP). Second, we consider a different approach to building CCA models which we call retrofitting. Retrofitting takes a model that does not attribute credit, and adds credit onto it. We prove a lower bound for CCA retrofitting under a weak optimality requirement. Given black-box access to the starting model, retrofitting requires query complexity exponential in the length of the model's outputs.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 15 of its 21 references that are in the map. First posted on arXiv 2 May 2026.

Builds on

15 of 21 references.

References · 15

  1. 2016Deep Learning with Differential Privacycited by 2,682
  2. 2019Data Shapley: Equitable Valuation of Data for Machine Learningcited by 617
  3. 2016Concentrated Differential Privacy: Simplifications, Extensions, and Lower Boundscited by 339
  4. 2016Concentrated Differential Privacycited by 139
  5. 2023On Provable Copyright Protection for Generative Modelscited by 97
  6. 2023Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generationcited by 94
  7. 2018Differentially Private Contextual Linear Banditscited by 82
  8. 2020Private Reinforcement Learning with PAC and Regret Guaranteescited by 51
  9. 2018Privacy-preserving Predictioncited by 48
  10. 2022Differentially Private Decoding in Large Language Modelscited by 34
  11. 2024Private prediction for large-scale synthetic text generationcited by 24
  12. 2024RAG with Differential Privacycited by 18
3 more
  1. 2025Source Attribution in Retrieval-Augmented Generationcited by 7
  2. 2023Private Everlasting Predictioncited by 5
  3. 2024Credit Attribution and Stable Compressioncited by 3

Built on by

Not cited yet.