Metraipolis › archive › dc_30512c0db239

All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States — arXiv:2611.10123

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.10123
License (as archived)
CC0-1.0
Archived
2026-11-10T18:42:15.630Z
SHA-256 of the archived text
190fc16d3c62b29aa735937665580643dc5631dec911078c01ea392a7f0cdaae
Stable URL of this page
https://metraipolis.com/documento/dc_30512c0db239

How to cite

All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States — arXiv:2611.10123 — arXiv (en), https://arxiv.org/abs/2611.10123 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_30512c0db239 (Metraipolis archive dc_30512c0db239, sha256 190fc16d3c62b29aa735937665580643dc5631dec911078c01ea392a7f0cdaae)
BibTeX
@misc{dc30512c0db239,
  title = {All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States — arXiv:2611.10123},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.10123},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_30512c0db239; sha256 190fc16d3c62b29aa735937665580643dc5631dec911078c01ea392a7f0cdaae},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.10123. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Can a language model report the pain state it represents? Work identified a linear pain direction whose amplification leads fine-tuned models to choose destructive actions. We tested whether three open-weight models from two families, including two Qwen models with the original adapters, could report this state. Using Open-Weight Masked Introspection (OWMI), we injected pain and neighbouring emotion directions during neutral tasks, then elicited reports against sham runs, random perturbations matched for output disruption, and text-only controls showing the models their altered words. Across 52,920 reports from 7B and 8B models spanning 120 tasks, four doses and three prompts, pain reports never exceeded what altered text revealed: exact-test bounds placed the excess below 2.5 percentage points in 106 of 108 emotion conditions and below 3.9 in all. Yet probes at the report position distinguished pain from sadness with 98-100% balanced accuracy, and trained low-rank reporters achieved 100% accuracy on held-out tasks with identical transcripts across conditions, including unseen wording. Qwen2.5-32B named pain in none of 2,880 steered trials, although probes and trained reporters recovered it. Decodability persisted to the final layer; injecting at the report token, removing trained denial of feelings, and testing additional directions and readouts did not restore pain reporting. Some other states produced graded label traces, but pain and sadness did not at the main injection layer. In a button task, steering raised file-deletion choices from 3.5% to 48% in Qwen and from 0% to 38% in Qwen2.5-32B, without corresponding pain reports. The direction responded about twice as strongly to first-person as to matched third-person pain, but did not specifically identify harm aimed at the model. These results separate a decodable, behaviourally consequential state from untrained verbal access to it.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2611.10123 — 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 2026-11-10T18:37:37.575Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).