A probe decodes the tool-call dependency DAG that causally propagates

Sun & Kazakov train a logistic-regression edge probe on all 65 residual-stream layers of Qwen3-32B to recover the tool-call dependency DAG in tau-bench's retail split [sun-kazakov-2026-tool-call-dependency-structure-is-linearly-decodable-in-llm-agent-residual-streams] The probe reaches AUROC 0.869 on held-out pairs, above a random-label control (0.491), a positional-only baseline (0.792), a random-init model (0.738), and an n-gram decoder (0.830) [sun-kazakov-2026-tool-call-dependency-structure-is-linearly-decodable-in-llm-agent-residual-streams] Signal emerges sharply by layer 14 (~22% depth) and plateaus through the final layer [sun-kazakov-2026-tool-call-dependency-structure-is-linearly-decodable-in-llm-agent-residual-streams] Per-layer activation patching shifts the probe's prediction at later non-patched layers toward a donor trajectory, so the representation causally propagates [sun-kazakov-2026-tool-call-dependency-structure-is-linearly-decodable-in-llm-agent-residual-streams]