Compression Favors Consistency, Not Truth: When and Why Language Models Prefer Correct Information
This addresses the problem of understanding truth bias in language models for AI researchers, but it is incremental as it builds on existing compression theories with controlled synthetic data.
The paper investigates why language models sometimes prefer correct statements despite training on mixed-quality data, introducing the Compression-Consistency Principle that explains this as a side effect of compression pressure favoring internally consistent hypotheses, with experiments on synthetic math corpora showing up to 83.1% accuracy in preferring correct completions under random errors but near-chance accuracy with coherent incorrect rules.
Why do language models sometimes prefer correct statements even when trained on mixed-quality data? We introduce the Compression--Consistency Principle: next-token prediction favors hypotheses that allow shorter and more internally consistent descriptions of the training data. Truth bias emerges only when false alternatives are structurally harder to compress. We test this using small GPT-2-style character-level transformers (3.5M--86M parameters) on synthetic math corpora with controlled mixtures of correct and incorrect rules. In the random-error setting, models strongly prefer correct completions in paired evaluation: 83.1% accuracy at balanced data and 67.0% even when correct rules appear in only 10% of the corpus. Replacing random errors with a coherent but mathematically incorrect rule system largely eliminates the preference (near-chance accuracy). In a more natural-language-like synthetic world, the effect is weaker but still present (57.7%). Additional experiments show that embedding verification steps can restore preference for correctness even at small scale, while increasing the number of consistent rules produces a graded improvement in accuracy. Our results suggest that what appears as a "truth bias" is largely a side effect of compression pressure and preference for internal consistency, rather than an intrinsic drive toward truth. Full code and data are available at https://github.com/Rai220/compression-drives-truth.