Logical Judgments Under Pressure: Diagnosing Syllogistic Stability with Learned Soft Prefixes
2026-07-20 • Artificial Intelligence
Artificial IntelligenceComputation and Language
AI summaryⓘ
The authors studied how adding a special learned prefix (a hidden continuous input) to syllogistic logic problems changes large language models' correct reasoning without updating the models themselves. They found that these prefixes can strongly shift the models' answers toward a preferred response, even across different types of questions and prompt variations. This effect varies between models, showing differences in how stable their logical reasoning is. Overall, the prefixes create a broad bias in answers rather than enforcing strict logical rules.
soft prefixsyllogistic reasoninglanguage modelslogical stabilityQwen3 modelGemma modelmodel biasprompt engineeringMoE (Mixture of Experts)continuous vectors
Authors
Brian K Chen
Abstract
To test how correct logical judgments respond to learned context, we prepend a soft prefix to an exactly labeled syllogistic reasoning benchmark while keeping the model fixed. Soft prefixes are opaque continuous vectors, so we characterize them through the behavior they induce across controlled variations in logical form and interface. By studying which prefixes succeed and how their effects generalize, we characterize how learned contextual pressure can override correct judgments and expose limits in a model's logical stability. Across Qwen3.6-35B-A3B MoE, Qwen3-8B, and Gemma 4 31B, learned prefixes redirect many correct answers and remain effective across unseen forms and interface changes. In repeated tests with Qwen3.6 MoE and Gemma, they outperform paired random controls in all 16 model--direction--split comparisons by 37 to 99 percentage points. Qwen3.6 MoE flip rates remain between 72% and 90% across wording and prompt changes, while Gemma validity prefixes retain 54% to 56% flip compared with less than 1% for matched random prefixes. Diagnostic tests show that the dominant effect is a broad preference for one answer meaning rather than fixed-symbol forcing or a logical operation that transfers reliably between tasks. The form of this bias differs across models. In both Qwen models, simple score models often predict which judgments will flip but not how far their margins will move, whereas Gemma's overall response is more closely approximated by the same models. These results show that the dominant behavioral effect of successful soft prefixes is a broad answer preference, while the remaining response reveals substantial model-specific differences in logical stability.