DigestAI news desk

AI news, digested. Every story with its sources, every 30 minutes.

Researchupdated

Study finds causal control in subliminal prompting varies by model depth

Researchers from arXiv cs.CL present a study on subliminal learning, where language models transmit hidden traits through seemingly unrelated outputs. The paper challenges the common explanation of "token entanglement" by distinguishing between four different measurements: output co-variation, fixed vector alignment, state readability, and causal control. Using a fixed animal-number prompting…

1 source primary source

Key points

  • Fixed output-vector similarity predicts behavior less well in larger Llama-3.1 models, with a correlation change of -0.080.
  • Donor-control AUC rises from 0.254 to 0.540, showing causal control increases with depth in Llama-3.1.
  • Per-token averaging in Qwen models creates a length confound that masks the true one-token association.

The results show that fixed output-vector similarity predicts behavior less effectively as model size increases, with a paired mean correlation change of -0.080. However, causal control measurements reveal a different pattern. When copying temporary answer-position states between prompts, the donor-control AUC rises from 0.254 to 0.540, a significant increase of +0.286. This causal control remains strong even with eight transformer blocks remaining, suggesting that the mechanism is not simply a matter of static geometry.

The study also examines Qwen models and finds that scoring every digit in sequence does not recover the positive one-token association. Instead, per-token averaging creates a false positive association that disappears when controlling for number width, indicating a length confound. The authors conclude that fixed geometry, observational readability, causal timing, and multi-token measurement are distinct properties. While these findings constrain token-level explanations, they do not yet identify the exact mechanism of training-time trait transfer.

Read the original atarXiv cs.CL · by Barath Velmurugan primary sourceOpen source ↗
Topics · follow one to build your own front page
Llama-3.1-8BLlama-3.1-70BQwen

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us.

Comments

via GitHub Discussions

More in Research

All →

Related stories