Metraipolis › archive › dc_678ec91e25ea
When Steering Shifts Discourse: Identifying and Intervening on Discourse-Level Changes due to Activation Steering in LLMs — arXiv:2610.41661
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.41661
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-02T07:01:13.017Z
- SHA-256 of the archived text
5a5595a7f31d91d6bcf14802cd39dc13b0d63b0bfe9b97502db338767a8dc691- Stable URL of this page
https://metraipolis.com/documento/dc_678ec91e25ea
How to cite
When Steering Shifts Discourse: Identifying and Intervening on Discourse-Level Changes due to Activation Steering in LLMs — arXiv:2610.41661 — arXiv (en), https://arxiv.org/abs/2610.41661 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_678ec91e25ea (Metraipolis archive dc_678ec91e25ea, sha256 5a5595a7f31d91d6bcf14802cd39dc13b0d63b0bfe9b97502db338767a8dc691)
BibTeX
@misc{dc678ec91e25ea,
title = {When Steering Shifts Discourse: Identifying and Intervening on Discourse-Level Changes due to Activation Steering in LLMs — arXiv:2610.41661},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.41661},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_678ec91e25ea; sha256 5a5595a7f31d91d6bcf14802cd39dc13b0d63b0bfe9b97502db338767a8dc691},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.41661. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Activation steering is an inference-time technique used to control or modify the behavior, tone, or safety properties of a large language model (LLM) without retraining the model's weights or editing its input prompt, but simply by adding a vector to its residual-stream activations. However, activation steering also changes the generated text in unintended ways, unrelated to intended behavior change, that are not well understood. Prior work has measured these changes only through coarse signals such as perplexity. In this work, we show that steering systematically produces significant changes in discourse dimensions of the generated text, such as its formality, use of evidence, and ambiguity. We present a way of countering this by adding a second, complementary vector that targets the affected discourse dimension. This intervention restores about 78% of the unintended discourse changes while preserving the intended steering behavior in about 93% of the generations. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.41661 — 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 2026-11-02T06:56:58.402Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).