Metraipolis › archive › dc_bc01807c4fca

Reading the Gate, Not the Interference: Output-Side Interference Measurement Does Not Track Merge Collapse — arXiv:2608.11797

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.11797
License (as archived): CC0-1.0
Archived: 2026-10-19T05:39:55.872Z
SHA-256 of the archived text: 2067c0a024bf75e864f8c2b8adccf18c40ff51dd32ed45af882bf8668fb9a994
Stable URL of this page: https://metraipolis.com/documento/dc_bc01807c4fca

How to cite

Reading the Gate, Not the Interference: Output-Side Interference Measurement Does Not Track Merge Collapse — arXiv:2608.11797 — arXiv (en), https://arxiv.org/abs/2608.11797 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bc01807c4fca (Metraipolis archive dc_bc01807c4fca, sha256 2067c0a024bf75e864f8c2b8adccf18c40ff51dd32ed45af882bf8668fb9a994)

BibTeX

@misc{dcbc01807c4fca,
  title = {Reading the Gate, Not the Interference: Output-Side Interference Measurement Does Not Track Merge Collapse — arXiv:2608.11797},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.11797},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bc01807c4fca; sha256 2067c0a024bf75e864f8c2b8adccf18c40ff51dd32ed45af882bf8668fb9a994},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.11797. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Task-arithmetic merging works until it doesn't, and the field diagnoses why by measuring interference inside the merged model. We take the most direct such measure, the exact layerwise activation cross-term of a factorial ledger, establish its causal anatomy, and then ask what it tracks. The anatomy is clean: each block mostly transports and amplifies the cross-term rather than generating it; erased, it is regenerated by the untouched marginal paths to 99% of its norm unless removed late; its output effect varies monotonically with the displacement's angle (orthogonal displacements make interference worse), and a two-assumption model derives the angle law and retro-dicts the dose curve (R^2 >= 0.99). What the measure tracks is not what the field assumes. Behavioural expert-likeness is decoupled from it across four instruments. Its cross-condition behaviour is denominator-dominated: an instruction template pins the main effect to within 1% while the absolute interaction grows 111x from two to six merged tasks, suppressing expressed interference at k=2 and amplifying it at k=6. And where merging actually collapses, the cross-term is a bystander, not the carrier: across two collapse parameterizations at two scales, even erased persistently at every position, removing it entirely repairs none of the collapse. There the output-side ratio carries no method information under a common counterfactual, while two state-space measures the field already uses rank methods correctly at both scales. All 81 predictions were frozen before their data; falsifications are reported as such. Output-side interference measurement reads the gate, the denominator, and the displacement budget, not the interference. What fails a merge is the carrier-bystander split: collapse rides in the marginal displacements while the cross-term merely accompanies it, and only state space sees the carrier.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-13 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.11797 — 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 2026-10-19T05:18:32.489Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).