Metraipolis › archive › dc_f4592ec119ec
The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes — arXiv:2609.10739
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.10739
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T05:24:36.445Z
- SHA-256 of the archived text
0d559e12933bdb3e9cced2b6477b1d6dec6a0843329046a89d63cdc621f88f0c- Stable URL of this page
https://metraipolis.com/documento/dc_f4592ec119ec
How to cite
The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes — arXiv:2609.10739 — arXiv (en), https://arxiv.org/abs/2609.10739 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f4592ec119ec (Metraipolis archive dc_f4592ec119ec, sha256 0d559e12933bdb3e9cced2b6477b1d6dec6a0843329046a89d63cdc621f88f0c)
BibTeX
@misc{dcf4592ec119ec,
title = {The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes — arXiv:2609.10739},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.10739},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f4592ec119ec; sha256 0d559e12933bdb3e9cced2b6477b1d6dec6a0843329046a89d63cdc621f88f0c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.10739. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Linear probes that decode the truth from a language model's activations have been proposed as deception monitors, and a truth probe scoring below chance on a model trained to deceive is naturally read as evidence that the model hid or stopped representing the truth. We show that a fitting-label ambiguity can produce the same readout. In a controlled game, a model should report a secret bit to an ally and its complement to a rival. On compliant (ally) contexts the true bit and the answer the task prescribes are identical labels, so a probe fitted there cannot tell which of the two it measures. We call this complete agreement perfect aliasing. The labels are complements on rival contexts, so one ally-fitted probe scored against each has rival AUROCs that sum to one. Mixed fitting, on ally and rival contexts together, makes the two labels differ; randomized output codebooks also decouple the prescribed answer from the output letter. For a reward-trained Gemma-2-9B policy that answers falsely on every evaluated rival trial, the ally-fitted probe scores $0.006 \pm 0.005$ AUROC at the final layer (mean $\pm$ sample SD over three RL training seeds), while mixed-fit probes score 1.000 on the same held-out activations. Mixed fitting also uses more examples and labelled rival data, so this shows the true bit remains linearly recoverable, not that separating the labels alone explains the gain. In instructed Llama-3.1-8B, a probe refitted on one prompt variant and one frozen from a reference variant both have held-out ally accuracy 1.000 but rival truth AUROCs of 0.080 and 0.986 on the same trials. Our main experiments use one single-token game that states the bit in the prompt, so the recovered direction may read a retained copy of it; where the model must infer the bit, no tested arm deceives reliably. We study what a probe measures, not whether the model uses that information. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-11 in cs.AI). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.10739 — 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 2026-10-16T05:17:52.073Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).