Logical Judgments Under Pressure: Diagnosing Syllogistic Stability with Learned Soft Prefixes
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
Reading in Plain English mode — toggle below to switch
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
This research paper advances MoE and Reasoning, directly affecting how engineers build, optimize, and ship AI systems. Teams running production LLM workloads need to understand these techniques to stay competitive in hiring and architecture decisions.