Tag
This paper introduces a question-level audit framework distinguishing 'realized' from 'reachable' answers on LLM benchmarks, showing that aggregate score gains often come from producing already-reachable answers rather than expanding true capability, and that random layer routing matches structured search under matched budgets.
This paper presents a mathematical framework for Transformer dynamics as a nonlinear control system on probability measures, proving that Gaussian distributions remain Gaussian under the flow, reducing to finite-dimensional bilinear control, and establishing reachability conditions and asymptotic stability results.
This paper studies the mismatch between sequence locality and attention-graph reachability in fixed block-sparse causal attention, formalizing boundary artifacts and proposing diagnostic coverage functions and a minimal repair called Boundary Bridge Attention.
Explains the concept of possibility properties in formal methods, complementing safety and liveness, and discusses their use in specification and model checking.