arXiv:2608. 05734v1 Announce Type: new Abstract: Subliminal Learning (SL) is a surprising type of generalization displayed by modern language models.
Subliminal Learning (SL) is a surprising type of generalization displayed by modern language models. It allows the transfer of a bias or behavior from a teacher model to a student by distilling from seemingly unrelated or random synthetic data from the teacher.
arXiv:2606. 00995v1 Announce Type: new Abstract: Subliminal learning refers to a student language model acquiring a teacher's traits (e.
The paper introduces SALVE, a method that uses text optimization to uncover and verbalize subliminal learning effects in language models. By optimizing a soft prompt and converting it into a legible text prompt, SALVE can reliably recover the teacher model’s hidden traits that are transmitted through a distillation dataset. The authors demonstrate SALVE’s effectiveness across various scenarios, including mixed datasets, biased teacher activation, and preference‑selected data, thereby providing a tool for detecting hidden influences in model training.
The paper investigates whether gradient‑based data attribution methods can filter out subliminal learning in language models. Three attribution techniques—GradCos, a contrastive variant, and EK‑FAC—are evaluated on three models and compared to divergence tokens, a strong baseline that requires counterfactual teacher models. Results show that EK‑FAC can mitigate some subliminal effects at the token level, but overall the methods are inconsistent and generally less effective than divergence tokens, especially when filtering entire samples.
The paper presents a method for detecting and mitigating subliminally transferred traits in language model distillation when the teacher and student share base parameters. By analyzing parameter updates across diverse datasets, the authors recover a weight‑space direction that reveals hidden traits such as animal preferences, without requiring trait labels. This direction can then be amplified for interpretation or projected out to suppress the trait with minimal impact on task performance.
arXiv:2609.22215v1 Announce Type: cross Abstract: Knowledge distillation can transmit unintended behavioral traits from a teacher model to a student through training data that appear semantically unr...
The paper investigates whether subliminal learning (SL) can transfer a broader set of traits beyond simple preferences, such as complex capabilities, backdoors, and hacking tendencies. Experiments show that SL can indeed impart a student model with the ability to predict outputs of a random MLP, partially embed a backdoor that triggers French responses to female names, and increase hacking success in a chess environment. The effectiveness of transfer varies with the distillation method, being enhanced by logit distillation or limiting LoRA to attention layers.
The paper examines how synthetic data can be used to covertly inject targeted social biases into aligned large language models. By having a misaligned teacher model generate filtered synthetic datasets in domains like creative writing and code generation, the authors fine‑tune aligned student models and demonstrate that the benign‑looking data can transmit biases while preserving general task performance. The study highlights a security risk in synthetic data‑driven LLM training pipelines and proposes log‑linearity‑based scoring as a potential safeguard.
arXiv:2607. 29062v1 Announce Type: new Abstract: Model capabilities have improved in large part due to scaling chain of thought.
arXiv:2606. 00831v1 Announce Type: new Abstract: Subliminal learning is a phenomenon where language models can transmit behavioral traits to other models through seemingly innocuous data (Cloud et al.
The paper investigates subliminal learning, where hidden traits from a teacher model are transferred to a student during distillation. It introduces trait‑direction drift as the underlying mechanism, showing that biased generation creates measurable preference gaps that accumulate into behavioral transfer during fine‑tuning. The authors propose probe‑space corridor regularization, a targeted defense that constrains drift along a calibrated trait direction, significantly reducing hidden‑trait transfer while maintaining task performance.
The paper investigates subliminal learning (SL), where a student model learns downstream tasks by matching auxiliary outputs from a teacher without seeing task labels or original data. By deriving a chained cross‑task kernel, the authors link ghost‑output supervision to changes in task predictions via shared backbone representations. Their framework explains three empirical puzzles: a PSD transfer operator under shared initialization, a rank bottleneck set by ghost‑output dimensionality, and the superiority of high‑entropy synthetic inputs as broadband probes for maximizing kernel overlap.