Sender Continuation Perplexity: Measuring Reasoning Trajectory Alignment at the Natural Language Boundary

Zenodo (CERN European Organization for Nuclear Research) 2026 Open access

Read full text
Authors: Jeremy McEntire
DOI: 10.5281/zenodo.18843841
Cited by: 0

Abstract

Text preserves 63.7% of reasoning trajectory (PPL 3.05 vs original 1.59). Full activation without text scaffold achieves PPL 5.60, indistinguishable from no context (5.62). Despite 4x geometric fidelity, activation injection without text carries zero reasoning trajectory information. The 36.3% gap is the information ceiling for text-based coordination.

Fields of Study & Subjects

Topic Modeling Natural Language Processing Techniques Logic Reasoning and Knowledge

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
2 March 2026
Language
EN