Artificial Intelligence · 31.07.2026, 06:33 UTC
Subtract or Replay? Exact Deletion from Language-Model Memory
| Schweregrad | info |
|---|---|
| Kategorie | Artificial Intelligence |
| Quelle | arXiv cs.CL ↗ |
| Veröffentlicht | 31.07.2026 UTC |
Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.
arXiv:2607.27539v1 Announce Type: cross Abstract: Exact deletion from persistent language-model memory depends on how that memory represents a record. Addressable influence can be removed by algebraic decrement; influence transformed by later writes inside shared recurrent state requires rebuilding from before the write. We test this distinction in two pretrained models against explicit record-omitted references. First, we replace Gemma 3's global-attention layers with support-vector memory. After low-rank recovery at 1B, decrement and retained-key refit agree at the next-token output to median KL $5.4\times10^{-15}$ over 31 support-token deletions, with $+2.0\%$ perplexity relative to a matched fine-tune. A masked-refit proxy is indistinguishable from the never-ingested floor under elicitation, relearning, sampling, and LiRA attacks. At 4B and 12B, certificate ordering persists but utility cost rises to $11.2\%$ and $44.3\%$. Second, in a 48B Kimi Linear hybrid, additive writes admit a fixed decrement and diagonal decay a corrected one, whereas the delta rule makes $12$--$49\%$ of a record's contribution suffix-dependent. Checkpointed rewind-and-replay deletes real clinical records at contexts up to 18,842 tokens, matching never-ingested logits and all recurrent states bit for bit within a deterministic MLX implementation; replaying a correction provides exact amendment. Exact deletion is therefore a property of memory representation: subtract addressable records and replay entangled writes.