Metraipolis › archive › dc_3f3da78cab65

How Language Models Choose Sides: Internal Representations of Instruction Hierarchy — arXiv:2608.28648

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.28648
License (as archived)
CC0-1.0
Archived
2026-10-17T07:41:45.810Z
SHA-256 of the archived text
50dc2b32da5a10e3c490459c86c6be2175621a536c09172f264325c738a1a4d1
Stable URL of this page
https://metraipolis.com/documento/dc_3f3da78cab65

How to cite

How Language Models Choose Sides: Internal Representations of Instruction Hierarchy — arXiv:2608.28648 — arXiv (en), https://arxiv.org/abs/2608.28648 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3f3da78cab65 (Metraipolis archive dc_3f3da78cab65, sha256 50dc2b32da5a10e3c490459c86c6be2175621a536c09172f264325c738a1a4d1)
BibTeX
@misc{dc3f3da78cab65,
  title = {How Language Models Choose Sides: Internal Representations of Instruction Hierarchy — arXiv:2608.28648},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.28648},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3f3da78cab65; sha256 50dc2b32da5a10e3c490459c86c6be2175621a536c09172f264325c738a1a4d1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.28648. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We study how instruction-tuned LLMs arbitrate direct conflicts between system and user instructions. We introduce a benchmark of 41 paired constraints with deterministic verifiers and evaluate eight models under matched baseline, conflict, and same-channel control conditions. Behaviourally, the models split into three regimes by System Authority Delta: hierarchy-respecting models use the system channel as an authority signal, anti-hierarchy models follow the system less often than their same-channel baseline predicts, and no-effect models show little channel sensitivity. Llama-3.1-8B is the strongest anti-hierarchy case in our suite, following the system in only 0.10 of conflict trials. We use this behavioural failure case to ask whether user-preferring arbitration reflects the absence of an internal conflictresolution signal. It does not: on Llama-3.1-8B, the conflict outcome is linearly decodable from residual-stream activations at 0.97 balanced accuracy, 17 percentage points above a metadata-only baseline, with analogous signals on Qwen2.5-7B and gpt-oss-20b. Steering with a layer-12 mean of four per-conflict logistic-regression directions raises genuine system compliance from 0.132 to 0.530, while directions selected mainly for pooled separability steer poorly. User-preferring conflict resolution can therefore coexist with a readable internal arbitration signal, and successful intervention depends on the geometry of the readout rather than probe accuracy alone

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.28648 — 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 2026-10-17T07:28:26.471Z 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).