Abstract: Information has expanded from a statistical quantity in communication theory into a cross-disciplinary concept used to describe physical structure, biological heredity, cultural transmission, and conscious experience. However, these domains differ in carriers, persistence mechanisms, observation scales, and validation criteria, making it difficult to unify them through a single definition. To address this problem, this study proposes a four-factor framework for unified information theory consisting of the Biton (B), Gene (G), Meme (M), and Yuanon (Y). The study adopts replicator analysis, mechanism comparison, and cross-level causal modeling, and compares the four factors through five shared analytical dimensions: structural retention, replication or recurrence, sources of variation, selection conditions, and causal effects. The Biton is defined as an information pattern in physical or cyber domains that preserves and transmits structural constraints and produces testable effects on subsequent system states. Physical bitons include structures maintained through template induction, process recurrence, or autocatalysis, whereas digital bitons rely on storage, computation, copying, synchronization, and version migration. Genes are treated as functional hereditary sequence units that produce phenotypic or regulatory effects; memes are culturally transmitted patterns that can be repeatedly identified, reconstructed, varied, and linked to behavioral or institutional effects; and yuanons are proposed as candidate experiential units that can be distinguished, combined, and reactivated within conscious experience. The comparison shows that the four factors share common analytical dimensions but are not isomorphic entities. Their carriers, persistence mechanisms, fidelity, selection environments, and observation scales differ substantially. Genes have relatively mature molecular and population mechanisms, memes generally persist through social reconstruction, bitons range from structural recurrence to engineered digital replication, and yuanons still require operationalization and empirical validation. On this basis, a cross-level causal network is constructed among B, G, M, and Y. The forward chain B→G→M→Y is decomposed into material realization, biological enabling, and experiential triggering. Reverse relations include Y→M through expression and externalization, M→G through niche construction and changes in selection environments, G→B through expression and structure formation, and B→Y through sensory transduction and neural realization. These relations operate at different timescales and therefore require mechanism-specific observations and interventions. The framework further incorporates cyberspace enhancement. Digital representation, automated search, high-fidelity copying, and rapid feedback can alter replication fidelity, variation search rate, and feedback frequency, thereby accelerating information generation and transformation while also introducing risks such as error propagation, bias amplification, infrastructure dependence, and privacy concerns. Finally, the four-factor framework is mapped onto cyber-physical-social-thinking (CPST) space. Physical space is primarily associated with physical bitons, cyber space with digital bitons, social space with memes, and thinking space with yuanons, while the same object may have representations and effects across multiple spaces. This mapping distinguishes information-bearing objects from their digital representations and measurement proxies. The framework therefore shifts the goal of unified information theory from seeking a single definition of information to constructing a decomposable mechanism-based research program. Future work should develop quantitative indicators for bitons, information-bottleneck models for cross-level transformations, reproducible identification procedures for yuanons, and multilayer CPST models for empirical validation and engineering evaluation across heterogeneous information systems and cross-domain application scenarios.