Operational Proto-Introspection in Looped Language Models: Process-Quality Taps, Executable Branching, and the Readout-Control Boundary

תקציר מקורי באנגליתarXiv:2607.18553v2 Announce Type: replace-cross Abstract: Can a language model read the quality of its ongoing computation, and can an external intervention turn that readout into better outcomes? We test both questions in a frozen 2.6B looped transformer, Ouro-RLTT. On GSM8K, a strict pre-answer probe excludes the answer region and gold value yet predicts success: hidden states plus surface features (length, log-probability) reach AUROC 0.797, versus 0.731 for surface features alone (increment +0.066; task-clustered 95% CI [+0.021, +0.112]; 170 tasks, 680 candidates). On Horizon Logic, the same within-domain comparison gives +0.141 (CI [+0.004, +0.290]; 56 tasks), positive but imprecise. Low-capacity taps also read task-disjoint branch survival and generated-branch correctness, while recu