Metraipolis › archive › dc_428d2d4d3cd1

Language Models Encode the Contextual Truth of Propositions — arXiv:2608.03035

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.03035
License (as archived)
CC0-1.0
Archived
2026-10-31T12:30:53.912Z
SHA-256 of the archived text
cd6dc5ecda9c7d540a2076131a9fe009c53228d465f36b5f37394f2a6e57a553
Stable URL of this page
https://metraipolis.com/documento/dc_428d2d4d3cd1

How to cite

Language Models Encode the Contextual Truth of Propositions — arXiv:2608.03035 — arXiv (en), https://arxiv.org/abs/2608.03035 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_428d2d4d3cd1 (Metraipolis archive dc_428d2d4d3cd1, sha256 cd6dc5ecda9c7d540a2076131a9fe009c53228d465f36b5f37394f2a6e57a553)
BibTeX
@misc{dc428d2d4d3cd1,
  title = {Language Models Encode the Contextual Truth of Propositions — arXiv:2608.03035},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.03035},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_428d2d4d3cd1; sha256 cd6dc5ecda9c7d540a2076131a9fe009c53228d465f36b5f37394f2a6e57a553},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.03035. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Prior work has shown that LLMs encode the truth of factual propositions along linear directions in activation space. It's unclear how these representations extend to contextual truth: propositions whose truth is determined by in-context evidence rather than world knowledge. We show that LLMs maintain a linear representation of contextual truth that persists across structurally different output policies, even when the output doesn't require the model to determine a proposition's truth, and show causal evidence via steering experiments. Using the transcripts from a collaborative vision-language task that requires two LLMs to maintain a shared common ground, we show that truth representations of a proposition are significantly swayed by partner assertions about that proposition, even when the LLM has enough evidence to determine its truth. We find evidence that propositions near the decision boundary are more susceptible to having their truth shifted through partner assertions. Separating representation from output distinguish two forms of sycophancy that output behavior alone cannot: the model may accommodate a false proposition while continuing to represent it as false, or shift its representation across the boundary. The latter is 2.59x more common when the model agrees by restating the false claim explicitly than when it agrees implicitly.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.03035 — 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 2026-10-31T12:17:37.732Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).