Metraipolis › archive › dc_8dfdacf84c90

Circuit-Diff: Factual Edit-based Intervention Method for Localizing Knowledge in Attribution Graphs — arXiv:2609.23892

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.23892
License (as archived): CC0-1.0
Archived: 2026-10-09T07:07:25.594Z
SHA-256 of the archived text: 11e7d02697966d6633c8797deea1c29757427f6d471aea39c9b3f27b6dfcb43c
Stable URL of this page: https://metraipolis.com/documento/dc_8dfdacf84c90

How to cite

Circuit-Diff: Factual Edit-based Intervention Method for Localizing Knowledge in Attribution Graphs — arXiv:2609.23892 — arXiv (en), https://arxiv.org/abs/2609.23892 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8dfdacf84c90 (Metraipolis archive dc_8dfdacf84c90, sha256 11e7d02697966d6633c8797deea1c29757427f6d471aea39c9b3f27b6dfcb43c)

BibTeX

@misc{dc8dfdacf84c90,
  title = {Circuit-Diff: Factual Edit-based Intervention Method for Localizing Knowledge in Attribution Graphs — arXiv:2609.23892},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.23892},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8dfdacf84c90; sha256 11e7d02697966d6633c8797deea1c29757427f6d471aea39c9b3f27b6dfcb43c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23892. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Mechanistic interpretability defines features as the fundamental units of a neural network and circuits as the weighted subgraphs that carry out its computation. Because individual neurons are polysemantic, Cross-Layer Transcoders (CLTs) were introduced as a way to approximate a model's circuits by generating an attribution graph. The nodes of that graph, however, are unlabeled features: reading a graph means pruning it and then working out by hand what each surviving node means. To make CLTs easier to use for circuit discovery, we introduce Circuit-Diff, which intervenes on the model itself with a low-rank factual edit and takes the features whose role in the attribution graph changes under that edit as related to the edited knowledge. On the edits we examine, the flagged nodes are not only detectors of the object token: read off the CLT's released feature dashboards, they include features for the history, geography and associations surrounding the old and new objects. We formalize the method, measure how reliable a frozen CLT remains after a factual edit, test the selected nodes causally by patching them on up to 24 CounterFact edits, give a case study, and release an open-source implementation built on the circuit-tracer package, together with two further tools (multi-prompt aggregation and rule-based supernode labeling).

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.23892 — 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 2026-10-09T06:58:09.345Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).