This translation is considered to align with the principles of univocity, scientificity, conciseness, and coordination within the expert argumentation (demonstration) system and has a certain foundation for use in the current Chinese context. However, after reading related interpretations, I have formed a different understanding of this naming approach.

From a standardization perspective, this naming scheme offers comprehensibility and communicative advantages in the short term. However, when examined through dimensions such as computational ontology, information structure, multimodal evolution, and back-translation consistency, its long-term adaptability remains to be further validated. Against this backdrop, an alternative path worth considering—'FuYuan'—gradually demonstrates stronger structural consistency and cross-contextual stability.

I. The Misalignment of Definitions: 'Origin' Cannot Substitute for 'Essence'

Article Viewpoint (Chen Xilin, Researcher at the Institute of Computing Technology, Chinese Academy of Sciences): Token's initial role in AI is as a 'basic semantic unit of language,' so 'CiYuan' better fits its essence.

This judgment holds rationality (rationality) in a historical context, but in the current era of significant technological paradigm shifts, this mindset is essentially an 'academic exercise in futility.'

At the logical level of term definition, it is crucial to strictly distinguish between 'initial application scenarios' and 'structural essential attributes.'

Token indeed originated in natural language processing (NLP), but in the evolutionary path of AGI, it has long transcended the boundaries of language models, evolving into a foundational unit for unified processing of text, images, voice, and even physical signals. In modern computing systems, Token's true structural ontology is that of a 'discrete symbolic unit,' not a language unit of a single modality.

If named based on its 'initial role,' computers (Computer) should still be called 'electronic calculating hands' (due to their original function of replacing human calculators), and the Internet (Internet) should be called the 'Cold War military network.' The fatal flaw in this naming logic is that it only sees technologies as 'temporary roles' at specific historical moments while ignoring their 'physical ontology' that transcends eras.

Historical paths cannot be equated with essential attributes. Similarly, we cannot permanently confine Token to the narrow context of 'words' just because it was initially used for processing text.

Using 'initial application scenarios' to define foundational concepts essentially substitutes historical path dependency for structural ontological truth. While this definition may provide convenience for understanding in the early stages of technology, it will quickly become obsolete and a cognitive obstacle during the paradigm expansion stage of multimodal explosion. In contrast, 'FuYuan' directly aligns with the symbolic ontology of cross-modal computing, defining not Token's 'past' but its 'truth.'

II. The Limits of Analogy: When Explanation Becomes Definition, Deviation Begins

Article Viewpoint (Dong Yuxiao, Associate Professor, Department of Computer Science, Tsinghua University): Through analogies like 'word clouds' and 'bag-of-words,' discrete units in multimodal contexts can be understood as 'generalized words.'

Professor Dong Yuxiao's analogy aids understanding but should not replace definitions. While this approach offers some explanatory insight, elevating it to a naming basis may lead to conceptual misalignment.

Methodologically, analogies serve to lower comprehension barriers, while definitions delineate semantic boundaries. When 'word' is expanded to cover image patches, voice segments, vector representations (embeddings), and even broader perceptual signals, its original linguistic properties are continually diluted, and semantic boundaries become blurred. This 'analogy-driven' expansion path may maintain explanatory consistency in the short term but is prone to semantic drift over long-term evolution.

In terms of cross-modal expansion capabilities, caution is needed regarding the 'slip' from analogy to definition. In the context of term approval, the boundary between 'explanatory metaphors' and 'ontological definitions' must be distinguished to prevent the former from substituting for the latter.

A more intuitive comparison: In popular science contexts, we can liken a lightbulb to an 'artificial sun' to enhance intuitive understanding, but in scientific naming systems, we cannot rename the current unit 'Ampere' as 'light unit' based on this analogy. The former belongs to descriptive expression, while the latter involves a strict measurement system and standardized definition—the two cannot be mixed.

Similarly, terms like 'word clouds' and 'bag-of-words' are essentially descriptive or statistical metaphors that aid in understanding data structures or distribution patterns. Token, as a foundational measurement (measurement) unit in large models, is deeply embedded in computing cost billing, model training, and academic measurement systems. When its usage scale reaches daily hundreds of billions to trillions of calls, its naming carries not just explanatory functions but also foundational engineering and standardization significance. At this level, terms must align with their ontological properties rather than rely on analogical extensions.

If this analogical logic is further extended to naming, it implicitly assumes a dangerous premise: since people are already accustomed to understanding Token through 'words,' it is acceptable to continue using this analogy. However, this is essentially a continuation of path dependency—using the convenience of existing cognition to substitute for correcting the conceptual ontology. In this sense, such naming is closer to a 'linguistic romanticism' than a strict alignment with computational ontology.

We cannot demand discussions of 'electronic horses' in electric motors just because 'horsepower' contains 'horse.' Analogies can inspire understanding but cannot define standards.

In contrast, 'Fu' ( talisman ), as a more neutral concept, naturally possesses cross-modal adaptability and can cover various information forms such as text, images, and voice without requiring additional explanations. Therefore, a naming path centered on 'symbolic units' is closer to Token's structural essence at the definitional level. Under this logic, 'FuYuan,' as the corresponding translation, exhibits higher conceptual consistency and long-term adaptability.

III. The Cognitive Cost: When Semantic Anchors Create Systemic Misunderstandings

Article Viewpoint (Synthesized Expert Opinions): The term 'CiYuan' is concise, aligns with Chinese habits, and is easy to disseminate.

This judgment holds some rationality in terms of communication, but its implicit premise is that the public can accept cross-modal analogies of 'word.' However, analogy is essentially an expert cognitive tool, not a natural mode of understanding for the general public. For ordinary users, 'word' has a strong semantic anchoring effect—upon hearing 'word,' their intuitive association inevitably points to linguistic systems rather than other modalities like images, sounds, or actions. This cognitive path is not a technical issue but a stable structure at the level of cognitive psychology.

Building on this, when 'word' is expanded into a so-called 'generalized word,' it actually creates a deviation in user cognition. Users first form an intuitive understanding of 'word = linguistic unit' rather than the abstract concept of 'cross-modal symbolic unit.' Once this misunderstanding is established, all subsequent explanations become corrections to existing cognition rather than natural extensions of understanding.

For example, when media reports state that 'the model used 10 trillion CiYuan for training,' the public is likely to interpret this as 'reading a large amount of text,' overlooking the substantial inclusion of images, voice, and other modal data. Such misunderstandings are not isolated cases but systemic inducements arising from the term's own semantic anchoring.

In practical engineering contexts, this naming may also create friction in cross-disciplinary communication. When discrete units in visual or voice models are called 'words,' it not only risks semantic misunderstandings but also generates unnecessary linguistic conflicts across different fields. Multimodal systems require unity at the 'symbolic layer,' not an expansion of linguistic categories.

In contrast, 'Fu,' as a more abstract concept, has a slightly higher initial comprehension threshold but points more neutrally in its semantic direction, without pre-locking cognition at the linguistic layer. In the long term, it is more conducive to establishing a stable, unified cognitive framework, thereby reducing overall explanatory costs and providing a more stable cognitive foundation for multimodal unification.

The cost of naming does not occur at the moment of definition but at the moment of correction; once early naming forms semantic anchors, the cost of subsequent cognitive repair rises exponentially.

Experts can extend the boundaries of 'word' through analogy, but the public will not understand concepts through analogy. Naming is not for experts but for the cognitive system of an entire era.

IV. The Illusion of Univocity: When One Word Attempts to Carry Two Systems

Article Viewpoint (Nomenclature Approval Principles): 'CiYuan' aligns with the principle of univocity and helps resolve translation confusion.

Regarding term univocity, special attention must be paid to the systemic risks posed by 'one word with two meanings.' In scientific nomenclature approval, 'univocity' is a foundational principle. If a term requires contextual or additional explanations to distinguish meanings, its value as a standard component is already lost.

However, from the perspective of the existing academic system, this judgment still leaves room for further discussion. The term 'CiYuan' is already well-established in linguistics and natural language processing (NLP), where it has long corresponded to the English concept of Lemma—the canonical form of a word (e.g., 'be' as the lemma for 'is/am/are'). This usage has formed a stable consensus in linguistics and NLP textbooks and academic papers.

Against this backdrop, translating Token as 'CiYuan' could easily lead to semantic conflicts in specific expressions, resulting in disastrous outcomes.

For example, when describing 'lemmatizing a token' in NLP, the Chinese expression would become 'lemmatizing a CiYuan,' creating a structure where 'CiYuan' is both the operator and the operand. This not only increases comprehension costs but also introduces ambiguity in academic writing and information retrieval, making it difficult for readers to distinguish whether 'CiYuan' refers to the discrete unit being segmented or the canonical form of a word.

Conceptually, the two also have clear distinctions: Lemma emphasizes 'reduction' at the linguistic level, corresponding to standardized expressions after morphological variations, while Token emphasizes 'segmentation' in the computational process, corresponding to the smallest discrete unit when a model processes information. This difference between 'reduction' and 'segmentation' aligns with different dimensions of the semantic versus symbolic layers.

Therefore, when a term requires 'generalization' to cover multiple existing concepts simultaneously, its univocity effectively transforms into 'unity at the explanatory level' rather than 'stability at the semantic level.'

When a term requires explanations to maintain unity, its stability as a standard term often begins to waver.

In contrast, 'FuYuan' does not present semantic conflicts within the existing terminology system. On the one hand, it retains Token's ontological property as a discrete symbol; on the other hand, it avoids overlapping with the established translation of Lemma, demonstrating higher stability in semantic clarity and systemic consistency.

V. The Return to Ontology: Token is Fundamentally a 'Symbol,' Not a 'Word'

Article Viewpoint (General Explanation): Token is the smallest unit used to process text in language models.

This description is functionally valid but remains at the level of 'how it is used' without touching upon its ontological properties in computational theory. From the perspectives of information theory and computational theory, the basic objects processed by computing systems are not 'words' but 'symbols' (symbol).

This can be further understood from two levels:

On the one hand, from an information theory perspective, the essence of information lies in eliminating uncertainty, with its measurement unit being the bit (bit), and its carrier entity being discrete symbols. Symbols do not concern semantic content but relate only to probability distributions and encoding structures.

On the other hand, at the computational implementation level, large models do not 'read characters'; their processing objects are discrete index representations (IDs). Whether this ID corresponds to a Chinese character, an image patch, or an audio sample point, it participates in computations in a unified symbolic form.

Within this framework, Token's essence lies precisely at the 'symbolic layer,' not the 'semantic layer.' Symbols themselves do not carry semantics but serve as the fundamental carriers of encoding and computation.

Naming Token as 'CiYuan' introduces an implicit linguistic semantic-layer orientation, pulling this originally symbolic-layer concept back into a language-centered understanding path. While this naming may offer intuitiveness at the explanatory level, it easily blurs the boundary between 'symbolic computation' and 'semantic understanding' at the theoretical level.

In contrast, 'FuYuan' remains conceptually within the symbolic layer. On the one hand, it accurately reflects Token's computational property as a discrete symbol; on the other hand, it avoids introducing semantic features into the ontological definition, thus better aligning with the fundamental frameworks of information theory and computational theory.

From a broader perspective, as AI systems continue to evolve toward multimodality and general intelligence, naming foundational concepts in direct alignment with their mathematical and computational ontology will better facilitate the construction of stable, scalable cognitive systems. In this sense, a naming path centered on 'symbolic units' is not merely a linguistic choice but a consistent expression of computational essence, with 'FuYuan' being the natural correspondence within this framework.

Defining concepts from the symbolic layer aligns with computational essence; naming concepts from the semantic layer is closer to explanation than definition.

VI. The Rupture of Language: Mapping Failure in Back-Translation Mechanisms

Article Viewpoint (Synthesized Interpretations): 'CiYuan' has gradually formed a usage foundation in Chinese academic circles and possesses certain communicative advantages.

In cross-linguistic contexts, caution is needed regarding the systemic impacts of 'back-translation fracture (rupture).' Measuring the long-term viability of a scientific term depends not only on its expressive capacity in the Chinese context but also on its ability to achieve stable mapping within the international academic system. Ideal terminology should possess 'reversibility,' meaning it can achieve semantically consistent round-trip translations across different languages.

The above judgment reflects 'CiYuan's' acceptability in the local context, but from a cross-linguistic perspective, there is still room for further discussion. If a term only succeeds in a single linguistic system without forming stable international correspondences, it may introduce additional comprehension costs in academic exchanges.

Specifically, 'CiYuan' lacks a clear, unique back-translation path. When rendered back into English, it often leads to divergence (divergence) among multiple approximate concepts: for example, 'word unit' lacks rigorous academic definition, 'morpheme' corresponds to linguistic morphemes, and 'lexeme' points to lexemes. None of these concepts accurately cover Token's meaning in computational contexts and may instead introduce categorical shifts.

In contrast, 'Fuyuan' can be more naturally mapped to 'symbolic unit.' This concept has a clear theoretical foundation and stable usage in fields such as information theory, discrete mathematics, and multimodal representation, maintaining consistent semantic orientation across different contexts. Therefore, it is easier to establish a one-to-one mapping between Chinese and English.

From a practical perspective, once a term enters academic papers, technical documentation, and international communication scenarios, its back-translation capability directly affects expression efficiency and comprehension accuracy. If a term requires additional explanation for cross-language conversion, its long-term usage costs will continue to accumulate.

Therefore, in cross-language systems, the main issue with 'Ciyuan' lies in the instability of its mapping path, while 'Fuyuan' demonstrates higher certainty in semantic correspondence and conceptual consistency. Against the backdrop of increasing globalization in artificial intelligence, selecting terms with strong back-translation characteristics will better facilitate the construction of an open and interoperable academic and technical system.

The international reversibility of terms is essentially a key criterion for their long-term academic viability.

VII. The Misconception of Uniformity: Formal Consistency ≠ Structural Consistency

Article Perspective (Synthesizing Expert Opinions): 'Ciyuan' maintains stylistic consistency with terms like 'embedding' and 'attention,' being concise and abstract, aligning with the Chinese technical context.

Conclusion First: The unification of a terminological system should be based on 'conceptual isomorphism' rather than 'linguistic homomorphism.'

In the supporting arguments for 'Ciyuan,' a common rationale is that its expressive style aligns with terms like 'embedding' and 'attention,' being concise and abstract, fitting the Chinese technical context. This rationale captures the genuine need for unity in a terminological system, but the issue lies in the fact that—if unity remains only at the linguistic level rather than the structural level—it shifts from 'order' to 'illusion.'

'Embedding' and 'attention' have become stable terms because they correspond to clear computational structures: the former is vector mapping, and the latter is a weighting mechanism, with their names directly pointing to the computational essence. In contrast, 'Ciyuan' belongs to explanatory naming, with its rationality (rationality) depending on the analogical framework of 'broad terms.' Once divorced from explanation, this naming itself lacks self-consistent structural orientation.

This difference raises a critical issue: formal consistency with semantic drift.

The former reduces expression costs, while the latter ensures cognitive stability. If priority is given to 'linguistic homomorphism,' complexity does not disappear but shifts into a long-term cognitive burden. Only naming based on 'conceptual isomorphism' can maintain stability across contexts and multimodal evolution.

When 'embedding,' 'attention,' and 'Ciyuan' appear side by side, it is easy to form the illusion of 'conceptual co-location.' However, in reality, the first two are mechanisms, while the latter is an object; the first two have strict definitions, while the latter relies on contextual explanation. This structural misalignment plants hidden discontinuities in the cognitive system.

More importantly, when the naming of a foundational concept relies on analogy rather than structural definition, its impact does not remain confined to a single term but spreads throughout the entire terminological system. When subsequent concepts attempt to revolve around this naming, they will have to continuously rely on explanations to maintain consistency, leading to implicit structural misalignments.

In this sense, 'Fuyuan' provides an expression path closer to the underlying structure. It directly points to the fundamental object in computational systems—the symbol—without relying on analogical explanations, maintaining consistency across different contexts.

Terminology is not just a label but an entry point to cognition. Good terminology makes explanations gradually disappear, while poor terminology leads to increasing annotations. When foundational concepts deviate from structure, the terminological system can only rely on explanations to sustain itself rather than on self-contained definitions.

Conclusion

Essentially, the choice of terminology is not merely a linguistic issue but an early shaping of the cognitive structure within a field. Once naming deviates from its structural essence in the initial stage, the subsequent system can only sustain operation through continuous explanations, struggling to form a self-contained conceptual network.

In the process of artificial intelligence moving toward generalization and multimodal fusion, a term that can align with the computational essence and possess cross-contextual stability is more likely to become a long-term effective cognitive cornerstone. In this sense, the naming path centered on 'symbolic unit' demonstrates a more balanced adaptability in balancing technical essence and cognitive clarity.