# Subliminal Signals in Preference Labels

arXiv: https://arxiv.org/abs/2603.01204 · HTML: https://mlsanity.com/paper/2603.01204/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 15 of its 16 references that are in the map.
- First posted on arXiv 1 Mar 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Human values and AI feedback alignment](https://mlsanity.com/alignment/human-values-and-ai-feedback-alignment/)

## Builds on (most cited first, 15 of 15)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](https://mlsanity.com/paper/2312.09390/) (2023)
- [Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge](https://mlsanity.com/paper/2410.02736/) (2024)
- [Alignment faking in large language models](https://mlsanity.com/paper/2412.14093/) (2024)
- [Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2406.07791/) (2024)
- [Natural Emergent Misalignment from Reward Hacking in Production RL](https://mlsanity.com/paper/2511.18397/) (2025)
- [Secret Collusion among AI Agents: Multi-Agent Deception via Steganography](https://mlsanity.com/paper/2402.07510/) (2024)
- [Subliminal Learning: Language models transmit behavioral traits via hidden signals in data](https://mlsanity.com/paper/2507.14805/) (2025)
- [Split and Merge: Aligning Position Biases in LLM-based Evaluators](https://mlsanity.com/paper/2310.01432/) (2023)
- [Eliminating Position Bias of Language Models: A Mechanistic Approach](https://mlsanity.com/paper/2407.01100/) (2024)
- [Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs](https://mlsanity.com/paper/2512.09742/) (2025)
- [Improving LLM-as-a-Judge Inference with the Judgment Distribution](https://mlsanity.com/paper/2503.03064/) (2025)
- [Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer](https://mlsanity.com/paper/2509.23886/) (2025)
- [Identifying and Mitigating Position Bias of Multi-image Vision-Language Models](https://mlsanity.com/paper/2503.13792/) (2025)
- [Mitigating the Bias of Large Language Model Evaluation](https://mlsanity.com/paper/2409.16788/) (2024)

## Built on by (most cited first, 3 of 3)

- [Can Data Attribution Filter Out Subliminal Learning? Not Reliably](https://mlsanity.com/paper/2609.20027/) (2026)
- [Why Ghost Outputs Teach: A Kernel-Based Understanding of Subliminal Learning](https://mlsanity.com/paper/2609.23260/) (2026)
- [Post-Training Leaves Behavioral Shadows on Unrelated Decisions](https://mlsanity.com/paper/2609.29233/) (2026)

## Abstract (arXiv)

As AI systems approach superhuman capabilities, scalable oversight increasingly relies on LLM-as-a-judge frameworks where models evaluate and guide each other's training. A core assumption is that binary preference labels provide only semantic supervision about response quality. We challenge this assumption by demonstrating that preference labels can function as a covert communication channel. We show that even when a neutral student model generates semantically unbiased completions, a biased judge can transmit unintended behavioral traits through preference assignments, which even strengthen across iterative alignment rounds. Our findings suggest that robust oversight in superalignment settings requires mechanisms that can detect and mitigate subliminal preference transmission, particularly when judges may pursue unintended objectives.