Aur Shalev Merin · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28498 · open-access PDF
Kami menguji ini dengan digit yang disalin dari pertanyaan, dalam model yang dikelilingi Ouro-1.4B dan transformator standar Qwen2.5-1.5B. Membandingkan prompts yang berbeda hanya dalam digit itu menunjukkan bahwa itu ada pada token prompt terakhir, dalam kode yang dibagikan di antara pertanyaan (baca sekitar 0.9). Sebuah probe yang melihat satu prompt pada satu waktu, pengaturan biasa, nyaris tidak bisa membacanya: 0.36 di Ouro dan 0.19 di Qwen bahkan setelah 5.249 pertanyaan pelatihan.
Machine translation (Meta M2M100); the English original follows.
Probing studies of latent reasoning often read a model's state where the answer begins and take a failed probe as evidence that a value is absent. We test this with a digit copied from the question, in the looped model Ouro-1.4B and the standard transformer Qwen2.5-1.5B. Comparing prompts that differ only in that digit shows it is present at the last prompt token, in a code shared across questions (readout about 0.9). A probe that sees one prompt at a time, the usual setting, barely reads it: 0.36 in Ouro and 0.19 in Qwen even after 5,249 training questions. The digit's change is small rather
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28498, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange