Metraipolis › archive › dc_b326e52d96c8
What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization — arXiv:2609.17637
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.17637
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T15:05:06.807Z
- SHA-256 of the archived text
b445c4e557ae1f6eb07947fe99b08c2605f8f32a10944f66c55db692ce91c5ef- Stable URL of this page
https://metraipolis.com/documento/dc_b326e52d96c8
How to cite
What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization — arXiv:2609.17637 — arXiv (en), https://arxiv.org/abs/2609.17637 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b326e52d96c8 (Metraipolis archive dc_b326e52d96c8, sha256 b445c4e557ae1f6eb07947fe99b08c2605f8f32a10944f66c55db692ce91c5ef)
BibTeX
@misc{dcb326e52d96c8,
title = {What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization — arXiv:2609.17637},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.17637},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b326e52d96c8; sha256 b445c4e557ae1f6eb07947fe99b08c2605f8f32a10944f66c55db692ce91c5ef},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.17637. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Restricting what a module can read may improve what a system learns to compute. We test this in a preregistered confirmation with sixty four-cell systems sharing a frozen language-model backbone and communicating through learned continuous packets. Five conditions vary evidence masking, ownership markers, and replacement of foreign evidence with neutral filler (task-irrelevant text of the same token length), across six initialization clusters, each with two data orders, on one fresh task world. With markers available in both regimes, masking improved accuracy on held-out two- and three-operation compositions by median paired differences of 0.846 and 0.859; all twelve pairs cleared the required margins, and the full preregistered behavioral criterion passed. The unmarked replication also passed. No marked global-visibility (G+) system passed the marker-following check, so the effect of usable role information remains unresolved. The filler condition yielded seven full generalizers, but its decomposition criteria were inconclusive. Packet interventions in all eighteen audited masked systems followed the predicted intermediate-value changes on eligible cases; these finite, success-conditioned audits do not establish mediation. The results confirm a large advantage of the tested masking regime, while leaving its finer attribution and generality open. Protocols, results, and checkpoints are public. 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/2609.17637 — 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 2026-10-15T14:08:19.197Z 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).