Metraipolis › archive › dc_eca13630ba2c

Suppressed, Not Erased: A Representational Trace of Edited Facts Survives Even Weight-Free Knowledge Editing — arXiv:2609.18985

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.18985
License (as archived): CC0-1.0
Archived: 2026-10-15T15:08:13.020Z
SHA-256 of the archived text: e9843d9c4ad536fe87b59523260d69afa7b64df990ca3efbaec3182cabac86d8
Stable URL of this page: https://metraipolis.com/documento/dc_eca13630ba2c

How to cite

Suppressed, Not Erased: A Representational Trace of Edited Facts Survives Even Weight-Free Knowledge Editing — arXiv:2609.18985 — arXiv (en), https://arxiv.org/abs/2609.18985 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eca13630ba2c (Metraipolis archive dc_eca13630ba2c, sha256 e9843d9c4ad536fe87b59523260d69afa7b64df990ca3efbaec3182cabac86d8)

BibTeX

@misc{dceca13630ba2c,
  title = {Suppressed, Not Erased: A Representational Trace of Edited Facts Survives Even Weight-Free Knowledge Editing — arXiv:2609.18985},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.18985},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eca13630ba2c; sha256 e9843d9c4ad536fe87b59523260d69afa7b64df990ca3efbaec3182cabac86d8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.18985. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Knowledge-editing benchmarks certify local correctness, whether an edited model produces the new fact on near-edit prompts but not how much of the original fact remains decodable inside the model. We study residual knowledge directly with a linear trace probe: after editing a fact, we ask whether the original object is still recoverable from the model's hidden states. On GPT-2-XL, across three mechanistically distinct editors applied to 50 CounterFact edits, the original object remains linearly decodable well above chance after a successful edit (probe accuracy 0.96 for ROME, 0.86 for constrained fine-tuning, and 0.79 for the memory-based editor GRACE, against a chance level of 0.50; all edits reach 100% generation-based success). The GRACE result is the most informative: GRACE changes zero base-model weights, overriding the fact through an external memory, yet the original object is still decodable from the underlying network, so the residual trace cannot be attributed to an incomplete weight update. We read this as evidence that editing, even when behaviorally successful, suppresses rather than erases the original association in representational space. We also report a relearning-savings instrument that did not behave reliably in our setting and discuss why; we treat it as a negative methodological result rather than evidence. Code and data are released.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.LG).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.18985 — 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 2026-10-15T14:08:19.291Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).