Maybe it's honesty training: who really wrote the chapter, the model or itself?
jonasgeiping · x · 2026-10-23
Jonas Geiping follows up on his observation about Claude models' animism quirk: it might be honesty training seeping in — since the human didn't write the chapter, saying 'in this chapter we investigate' would be wrong. So who wrote it? 'Maybe it really did write itself.'
More from Models
- Open 14B Carbon-X reads raw DNA at 67% benchmark score, ~1000x cheaper than GPT-6.1 Sol — Thom_Wolf · 2026-10-23
- Marker's lite mode scores 75.1 vs 42.3 for Azure Read on OmniParseBench — VikParuchuri · 2026-10-23
- Datalab's Indra OCR: a 1M-parameter open model that hits 44 pages/s on CPU across 80+ languages — VikParuchuri · 2026-10-23
- Datalab lite OCR scores 97% on Armenian vs Azure Layout's 21%, benchmark shows — VikParuchuri · 2026-10-23
- Datalab lite: full-layout OCR at $0.50/1,000 pages, more accurate than Azure Read at half the price — VikParuchuri · 2026-10-23
- Datalab OCR tiers: lite at $0.50, balanced at $4.00, accurate at $10.00 per 1,000 pages — VikParuchuri · 2026-10-23