Tag
This paper introduces a trace-supervised symbolic neural CPU architecture that combines recurrent control, an explicit operation router, and masked register writeback to enable auditable, interpretable program execution, with quantization-simulated writeback preserving symbolic operation paths.
This paper proposes evaluating coding LLMs on their understanding of software execution beyond control flow, including predicting memory usage, runtime, and profiler outputs, finding that all tested models perform poorly, indicating a lack of deep software world model understanding.