Metraipolis › archive › dc_95481eaa1f63

Which Influence Are We Estimating? The Role of Counterfactual Specifications in Data Attribution — arXiv:2609.31214

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.31214
License (as archived)
CC0-1.0
Archived
2026-10-08T01:38:38.679Z
SHA-256 of the archived text
f8a667032664059cd431b4294de64da760a1c798d63e3626a0f6ce475045d17f
Stable URL of this page
https://metraipolis.com/documento/dc_95481eaa1f63

How to cite

Which Influence Are We Estimating? The Role of Counterfactual Specifications in Data Attribution — arXiv:2609.31214 — arXiv (en), https://arxiv.org/abs/2609.31214 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_95481eaa1f63 (Metraipolis archive dc_95481eaa1f63, sha256 f8a667032664059cd431b4294de64da760a1c798d63e3626a0f6ce475045d17f)
BibTeX
@misc{dc95481eaa1f63,
  title = {Which Influence Are We Estimating? The Role of Counterfactual Specifications in Data Attribution — arXiv:2609.31214},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.31214},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_95481eaa1f63; sha256 f8a667032664059cd431b4294de64da760a1c798d63e3626a0f6ce475045d17f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31214. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Estimating the influence of training examples on model behavior is essential for data debugging, valuation, and attribution. Existing influence estimators often produce incompatible rankings, which are commonly ascribed to approximation error. We argue that a more fundamental source of disagreement is specification mismatch: influence depends on the behavior being attributed, the intervention applied to each training example, and the counterfactual training process that maps the intervention to a model response. These choices are especially important when the target behavior requires a tractable surrogate, such as query loss, a logit, or a margin. We formalize influence as a counterfactual estimand, distinguish specification mismatch across estimands from approximation error in estimating a fixed estimand, and organize representative estimators by their implied specifications. We further derive a local decomposition that exposes how behavior signals, training signals, and counterfactual parameter responses interact. Controlled experiments show that exact estimands under different specifications can induce different rankings, whereas approximation error grows as perturbations move farther from their linearization points. Experiments on noisy label detection and LLM attribution show that specification choices significantly affect attribution quality, especially for the choice of behavior surrogate. Behavior-aligned specifications can identify target-specific training examples obscured by default loss-based or similarity-based specifications. These results establish specification analysis as a necessary first step for interpreting and comparing data influence estimators.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.31214 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T01:12:14.361Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).