https://ai-news.ypetit.net/sources/eleutherai/ Latest publications from EleutherAI, aggregated by AI News from the official source. en Tue, 10 Nov 2026 00:08:45 GMT https://blog.eleuther.ai/subliminal-learning-pretraining/https://blog.eleuther.ai/subliminal-learning-pretraining/Sun, 25 Oct 2026 00:00:00 GMTresearchCross-model distillation shows subliminal learning survives differences in initialization, data order, corpus, size, and architecture — and that transmitting and acquiring a preference are separable abilities. https://blog.eleuther.ai/nla-stability/https://blog.eleuther.ai/nla-stability/Tue, 22 Sep 2026 00:00:00 GMTresearchTesting how robust Natural Language Autoencoder verbalizations are to changes in training seed, paraphrasing, and targeted text edits. Work done during SOAR 2026, by Kaylee, Daniel and Abhinav, with author order randomized. https://blog.eleuther.ai/aletheia-retrospective/https://blog.eleuther.ai/aletheia-retrospective/Tue, 25 Aug 2026 16:35:00 GMTresearchWhat we learned while building black-box and white-box detectors for AI deception during Aletheia's Quest. https://blog.eleuther.ai/historical-books-ocr-leaderboard/https://blog.eleuther.ai/historical-books-ocr-leaderboard/Mon, 10 Aug 2026 00:00:00 GMTresearchTesting open OCR models on historical books with the FineBooks BHL OCR Leaderboard. https://blog.eleuther.ai/dynamical-models-of-ai-governability/https://blog.eleuther.ai/dynamical-models-of-ai-governability/Mon, 13 Jul 2026 00:00:00 GMTresearchA toy dynamical model of whether the AI workforce that builds future AI ends up cooperative or uncooperative: where the basin boundary lies, what current evidence says about which side we are on, and what would tell us we are on the good path. https://blog.eleuther.ai/reward-hacking-indicators/https://blog.eleuther.ai/reward-hacking-indicators/Wed, 15 Apr 2026 00:00:00 GMTresearchUsing importance sampling with fine-tuned donor prefills to predict reward hacking emergence during training