Metraipolis › archive › dc_4cd347ca862a

SKILL2STEER: One-Pass Hypernetwork for Mapping Natural-Language Skills to Low-Rank Post-Block Activation Shifts — arXiv:2611.04497

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.04497
License (as archived): CC0-1.0
Archived: 2026-11-04T21:51:37.746Z
SHA-256 of the archived text: b94d337c2c461f289bbdc130aec1052e4a1dfc6b159bbba0c41ccb422416b31c
Stable URL of this page: https://metraipolis.com/documento/dc_4cd347ca862a

How to cite

SKILL2STEER: One-Pass Hypernetwork for Mapping Natural-Language Skills to Low-Rank Post-Block Activation Shifts — arXiv:2611.04497 — arXiv (en), https://arxiv.org/abs/2611.04497 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4cd347ca862a (Metraipolis archive dc_4cd347ca862a, sha256 b94d337c2c461f289bbdc130aec1052e4a1dfc6b159bbba0c41ccb422416b31c)

BibTeX

@misc{dc4cd347ca862a,
  title = {SKILL2STEER: One-Pass Hypernetwork for Mapping Natural-Language Skills to Low-Rank Post-Block Activation Shifts — arXiv:2611.04497},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.04497},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4cd347ca862a; sha256 b94d337c2c461f289bbdc130aec1052e4a1dfc6b159bbba0c41ccb422416b31c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.04497. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Compiling a natural-language agent skill into model parameters removes the skill text from the prompt and leaves behind a small artifact that can be distributed and reused. Existing methods differ in where they inject the compiled quantity, and all of them assume that the artifact carries the content of the skill. We test that assumption. Through three controls on a retrieval agent, we find that a compiled skill shift carries a domain-general behavioural prior, not the content of the skill. A wrong skill document performs on par with the correct one, a single averaged bundle matches per-skill routing under a two-sided equivalence test, and a same-norm random direction collapses to the no-injection baseline, which shows that the trained direction stays load-bearing. We then introduce SKILL2STEER, a one-pass hypernetwork that reads a skill document once and maps it to low-rank activation shifts added at the output of every decoder block. Each skill compiles to 2.36M parameters, 9.25 times fewer than a weight-space adapter, while the base weights stay frozen and the generator runs only at compile time. The shift can be scheduled across retrieval steps and deployed as one routing-free bundle. On SearchQA, a seven-dataset benchmark, SKILL2STEER reaches 37.9 exact match, 9.8 points above the frozen model and 2.3 points above a weight-space method that ships 9.25 times more parameters per skill.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2611.04497 — 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 2026-11-04T21:38:49.714Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).