BAP - Budgeted Activation Propagation Engine
Abstract
BAP is a reasoning kernel that traverses a ThoughtSpace — a graph of atomic reasoning primitives — using a physics-inspired energy budget equation instead of generating tokens with an LLM. The core idea: reasoning is not generation. Reasoning is retrieval + propagation + pruning. The LLM is fuel, not the engine. Query → OOV Detection → SHSRS Index Search → Activation Scoring → Pruning → Path BAP produces a deterministic reasoning path in 4–6 steps for queries that would cost an LLM 240–360 tokens of chain-of-thought.
Fields of Study & Subjects
Topic Modeling
Information Retrieval and Search Behavior
Natural Language Processing Techniques
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 4 March 2026
- Language
- EN
- Document type
- Preprint