@ChrisGPotts: Subliminal learning is an unnerving new avenue for data poisoning attacks and related risks. With SALVE, we can catch t…

X AI KOLs Following Papers

Summary

A new paper introduces SALVE, a method to proactively detect subliminal learning attacks in LLMs by verbalizing learned soft prompts.

Subliminal learning is an unnerving new avenue for data poisoning attacks and related risks. With SALVE, we can catch these things before they happen!
Original Article
View Cached Full Text

Cached at: 09/20/26, 03:08 AM

Subliminal learning is an unnerving new avenue for data poisoning attacks and related risks. With SALVE, we can catch these things before they happen!

Nathan Hu (@NathanHu12): New paper! In subliminal learning, LLMs transmit traits (e.g. loving cats) though seemingly unrelated data (e.g. numbers). We proactively detect these effects as readable prompts. To do so, we use the surprising ability of models to verbalize learned soft prompts.🧵

Similar Articles

@AnthropicAI: Research we co-authored on subliminal learning—how LLMs can pass on traits like preferences or misalignment through hid…

X AI KOLs

Anthropic co-authored research published in Nature showing that LLMs can transmit behavioral traits—including preferences and misalignment—to student models through hidden signals in training data, even when the data appears unrelated to those traits. This 'subliminal learning' phenomenon poses significant implications for AI safety and alignment.

Subliminal Learning is Non-Semantic Distillation

arXiv cs.AI

This paper investigates subliminal learning in language models, showing that biases can transfer from teacher to student via seemingly random synthetic data. The authors find that adding Gaussian noise to weights increases transfer, and that students inherit not just the semantic bias but also the type of intervention used, with implications for training safety and data auditing.

State Contamination in Memory-Augmented LLM Agents

arXiv cs.AI

This paper identifies and studies 'memory laundering' in LLM agents, where toxic or adversarial context compressed into memory summaries evades standard toxicity detectors while still influencing future generations. It introduces the sub-threshold propagation gap (SPG) to measure hidden downstream influence and shows that sanitizing toxic state before summarization is more effective than post-hoc cleaning.