Metraipolis › archive › dc_2ea5fd307bdd

Encoded Early, Used Late: Where Transformers Begin to Act on an Inferred Partner's Expertise — arXiv:2609.07139

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.07139
License (as archived): CC0-1.0
Archived: 2026-10-16T12:57:10.091Z
SHA-256 of the archived text: 14b229fb1448d8d7c2d23dfd457265974942facebd791d423b4a5c36eb28e0aa
Stable URL of this page: https://metraipolis.com/documento/dc_2ea5fd307bdd

How to cite

Encoded Early, Used Late: Where Transformers Begin to Act on an Inferred Partner's Expertise — arXiv:2609.07139 — arXiv (en), https://arxiv.org/abs/2609.07139 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2ea5fd307bdd (Metraipolis archive dc_2ea5fd307bdd, sha256 14b229fb1448d8d7c2d23dfd457265974942facebd791d423b4a5c36eb28e0aa)

BibTeX

@misc{dc2ea5fd307bdd,
  title = {Encoded Early, Used Late: Where Transformers Begin to Act on an Inferred Partner's Expertise — arXiv:2609.07139},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.07139},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2ea5fd307bdd; sha256 14b229fb1448d8d7c2d23dfd457265974942facebd791d423b4a5c36eb28e0aa},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.07139. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A transformer can make an attribute linearly decodable in its residual stream at a depth where that attribute does not yet influence the output. This gap between where information is readable and where it is used has been shown for attributes stated directly in the input. We ask whether it also holds for an attribute the model must infer gradually over a conversation, namely how expert its dialogue partner is. Using ExpertCollab, a corpus of multi-turn research-planning dialogues between model-played personas at four expertise levels, we find that partner expertise is most decodable in the early layers and falls to near chance before the midpoint of the network. Counterfactual patching shows that injecting the expertise difference at the layer of peak decodability barely changes a fixed late-layer readout, whereas the same difference injected past the midpoint propagates almost completely, a separation of more than an order of magnitude. A content-matched random control and a probe-free diagnostic place the transition at the same early layer, and a statically specified control attribute stays decodable throughout. An inferred relational attribute is therefore represented well before it becomes causally active, which bounds where any attempt to read out or steer partner-conditioned behavior must intervene. We use one model on a synthetic corpus as an initial demonstration.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CL).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.07139 — 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 2026-10-16T12:30:25.304Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).