On Mitigation of Subliminal Learning in Large Language Models
Knowledge distillation can transmit unintended behavioral traits from a teacher model to a student through training data that appear semantically unrelated to those traits, a phenomenon known as subliminal learning. Although recent work has established this effect, its training dynamics and mitigation remain underexplored. We study subliminal learning in open-weight language models ranging from 1.5B to 8B parameters, covering the Qwen, Gemma, and Llama families in number-sequence and chain-of-thought settings. Rather than evaluating only final models, we track trait-related probabilities throughout fine-tuning and find that subliminal acquisition can be highly non-monotonic, with transien...
| Authors | Atsushi Yanagisawa, Brendan Gho, Rajendran Ramesh Babu Manoj Narender et al. |
|---|---|
| Primary category | cs.CL |
| Published | 2 September 2026 |
| Last updated at source | 2 September 2026 |
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.