Metraipolis › archive › dc_2cba6017a1ca
Knowing, and Saying It Only When Asked: LLM Endognostics and the Schizognosis of Minerva-7B — arXiv:2609.22219
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.22219
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:49:19.145Z
- SHA-256 of the archived text
93b780a0d51cc73e2abd59f185a579eac81d25cb26585c6b3dce0f49dab25cb8- Stable URL of this page
https://metraipolis.com/documento/dc_2cba6017a1ca
How to cite
Knowing, and Saying It Only When Asked: LLM Endognostics and the Schizognosis of Minerva-7B — arXiv:2609.22219 — arXiv (en), https://arxiv.org/abs/2609.22219 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2cba6017a1ca (Metraipolis archive dc_2cba6017a1ca, sha256 93b780a0d51cc73e2abd59f185a579eac81d25cb26585c6b3dce0f49dab25cb8)
BibTeX
@misc{dc2cba6017a1ca,
title = {Knowing, and Saying It Only When Asked: LLM Endognostics and the Schizognosis of Minerva-7B — arXiv:2609.22219},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.22219},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2cba6017a1ca; sha256 93b780a0d51cc73e2abd59f185a579eac81d25cb26585c6b3dce0f49dab25cb8},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.22219. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Evaluating an aligned language model by reading its answers assumes the answers carry the distinction the evaluator cares about. We introduce LLM endognostics, a white-box internal auditing framework designed to extract and causally manipulate latent knowledge within the residual stream. Applied to Minerva-7B-Instruct-v1.0 on 124 minimal prompt pairs over 12 categories of professional risk, behavioral evaluation fails on most of the set: the model acts identically on 63.7% of the pairs (95% CI [55.0%, 71.6%]), complying with or refusing both members. Yet, projecting the residual stream onto the vocabulary by a Jacobian lens reveals a statistically significant Contrastive Endognostic Margin, proving the model maintains robust risk differentiation internally. In a second protocol crossing 25 facts with five linguistic framings, we show that the model conforms to presupposed falsehoods in 72% of cases, despite representing the true entity in its latent layers. Surgically ablating the direction of the planted falsehood restores the correct answer in 11 of 25 suppressed cases (McNemar p = 0.0010), validated by blind human annotation (binary agreement kappa = 0.68). In contrast, an out-of-sample linear probe achieves 77% accuracy at layer 10, but its orthogonal ablation yields a 0% recovery rate. This establishes a fundamental theoretical dissociation: abstract linear representation does not imply causal control over verbalization. Our main contribution is the formalization of endognostics to prove that behavioral evaluation and internal reading systematically disagree in the common case, and that linear decodability is decoupled from causal control. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.22219 — 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 2026-10-09T06:48:44.896Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).