Metraipolis › archive › dc_a89302eb4522

Toward Latent Language Model Skills Steering and Optimization: An Empirical Study — arXiv:2608.29459

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.29459
License (as archived): CC0-1.0
Archived: 2026-10-17T07:44:19.892Z
SHA-256 of the archived text: 22edbd37f140db657fce44962bee3578b54d075cade2674664063518229ed4e6
Stable URL of this page: https://metraipolis.com/documento/dc_a89302eb4522

How to cite

Toward Latent Language Model Skills Steering and Optimization: An Empirical Study — arXiv:2608.29459 — arXiv (en), https://arxiv.org/abs/2608.29459 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a89302eb4522 (Metraipolis archive dc_a89302eb4522, sha256 22edbd37f140db657fce44962bee3578b54d075cade2674664063518229ed4e6)

BibTeX

@misc{dca89302eb4522,
  title = {Toward Latent Language Model Skills Steering and Optimization: An Empirical Study — arXiv:2608.29459},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.29459},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a89302eb4522; sha256 22edbd37f140db657fce44962bee3578b54d075cade2674664063518229ed4e6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29459. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Skills, as a useful abstraction for the procedural capabilities of large language models (LLMs), capture how models perform structured, multi-step reasoning and program execution. Existing approaches typically treat skills as explicit, surface-level constructs specified through prompts or programs, leaving open the question of how such procedural capabilities are represented inside the model and whether they can be manipulated as structured objects in latent space. In this empirical study, we investigate whether procedural LLM skills can be represented as directions in activation space and whether vector-space operations over these directions can express skill-level behaviors. We find that procedural skills admit a vector-space representation: individual skill directions can be activated to shift model behavior; independently extracted directions can compose to form higher-level skills. Contrastive directions yield context-conditioned algorithmic personalization and optimization trajectories over skill directions evolve non-monotonically, with intermediate states often surpassing fully optimized solutions. These results support a representation-level view of procedural LLM skills: they admit a latent vector-space organization that allows direct manipulation through internal interventions.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.29459 — 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 2026-10-17T07:28:26.553Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).