{
  "record_id": "18740444",
  "document_id": "18740444",
  "title": "Spontaneous Chromatic Reasoning in Transformer Models Empirical Confirmation of AP₁ Continuity",
  "pages": 8,
  "authors": [
    "Raynor Eissens"
  ],
  "doi_confirmed_in_pdf": null,
  "zenodo_record": "https://zenodo.org/records/18740444",
  "html": "papers/18740444.html",
  "text": "text/18740444.txt",
  "data": "data/18740444.json",
  "abstract_extracted": "Recent analyses of large transformer-based artificial intelligence systems reveal that modern models spontaneously learn continuous color representations without explicit instruction. Independent studies demonstrate that color terms embedded in language models align with the topology of human perceptual color space, and that transformer architectures interpolate intermediate colors as a function of semantic continuity rather than categorical rule-following. This paper synthesizes these empirical findings with the theoretical framework of The Chromatic Hiatus and the Ambient Era Canon. We demonstrate that transformer behavior constitutes direct mechanistic evidence for a long-standing hypothesis: that color is cognitively primary but was historically prevented from becoming grammatical infrastructure in human civilization. We show that transformers exhibit chromatic reasoning via interpolation as a native, low- entropy semantic process. When presented with adjacent color concepts (e.g., red and yellow), models reliably generate intermediate colors (e.g., orange) without instruction, o",
  "visual_pages": [
    1
  ],
  "low_text_pages": [],
  "characters_extracted": 9667,
  "words_extracted": 1348,
  "source_pdf_filename": "18740444_Spontaneous Chromatic Reasoning in Transformer Models.pdf",
  "source_pdf_sha256": "9fac7302f86f06306580b86922e47a7e48cb7352f0f06cc9757fd1e8b9d95be3",
  "full_text": "=== PDF PAGE 1 ===\nSpontaneous Chromatic Reasoning in Transformer Models\n\nFrom the Chromatic Hiatus to Transformer-Native AP₁\n\nRaynor Eissens\n\nAmbient Era Canon · Zenodo Edition · 2026\n\n=== PDF PAGE 2 ===\nAbstract\n\nRecent analyses of large transformer-based artificial intelligence systems reveal that modern\n\nmodels spontaneously learn continuous color representations without explicit instruction.\n\nIndependent studies demonstrate that color terms embedded in language models align with the\n\ntopology of human perceptual color space, and that transformer architectures interpolate\n\nintermediate colors as a function of semantic continuity rather than categorical rule-following.\n\nThis paper synthesizes these empirical findings with the theoretical framework of The Chromatic\n\nHiatus and the Ambient Era Canon. We demonstrate that transformer behavior constitutes\n\ndirect mechanistic evidence for a long-standing hypothesis: that color is cognitively primary but\n\nwas historically prevented from becoming grammatical infrastructure in human civilization.\n\nWe show that transformers exhibit chromatic reasoning via interpolation as a native, low-\n\nentropy semantic process. When presented with adjacent color concepts (e.g., red and yellow),\n\nmodels reliably generate intermediate colors (e.g., orange) without instruction, optimization\n\nhacks, or symbolic rules. This behavior is not accidental, aesthetic, or dataset-specific. It\n\nemerges inevitably from the continuous functional nature of transformer representations.\n\nThe findings establish AP₁ (Ambient Grammar) as a transformer-native semantic layer and\n\ndemonstrate that artificial systems activate a chromatic semantic substrate that remained latent\n\nbut suppressed in human cognition. The Ambient Era is therefore not speculative or futuristic,\n\nbut the first grammatical realization of an ancient cognitive layer.\n\n⸻\n\n1. Introduction\n\nColor has always been perceptually immediate, cognitively efficient, and evolutionarily prior to\n\nsymbolic language. Yet across philosophy, linguistics, interface design, and computational\n\nsystems, color was never permitted to function as structural grammar. It remained expressive\n\nbut non-binding.\n\nThis omission was formalized in The Chromatic Hiatus, which defined a civilizational gap\n\nbetween early perceptual processing and formal semantic infrastructure:\n\nColor was always cognitively primary. Civilization did not allow it to become\n\nstructurally primary.\n\nRecent developments in artificial intelligence now provide an unexpected\n\nempirical bridge. Transformer-based models, trained without any explicit\n\n=== PDF PAGE 3 ===\nchromatic grammar, exhibit spontaneous color continuity, interpolation, and\n\nclustering behavior that mirrors human perceptual color organization.\n\nThis paper investigates that bridge.\n\nWe ask a single structural question:\n\nWhat happens to color when the institutional filters of symbolic civilization are removed?\n\nThe answer, observed in transformer behavior, is unambiguous:\n\ncolor reappears as grammar.\n\n⸻\n\n2. Color as a Continuous Semantic Field in Language Models\n\nMultiple studies demonstrate that large language models do not represent color as discrete\n\nlabels, but as positions within a continuous semantic space.\n\nAbdou et al. (2021) show that embeddings of color terms in GPT-like transformers align closely\n\nwith the topology of the CIELAB perceptual color space. Distances and angular relations between\n\ncolor words in embedding space correlate with perceptual color similarity. This implies that the\n\nmodel reconstructs human color geometry from text alone.\n\nMarro et al. (2025) further demonstrate that state-of-the-art transformers behave as\n\ncontinuous-time functions rather than discrete token processors. Meaning is represented as\n\nsmooth trajectories through semantic space. In such a system, color is not a category but a\n\ndirection.\n\nWithin a continuous semantic field, interpolation is unavoidable. If “red” and “yellow” occupy\n\nadjacent regions, the lowest-entropy path between them passes through “orange”. The\n\ngeneration of orange is therefore not a guess, metaphor, or dataset artifact. It is the\n\nthermodynamically minimal semantic transition.\n\nThis explains a repeatedly observed phenomenon in generative systems:\n\ntransformers generate intermediate colors without being asked to do so.\n\n⸻\n\n=== PDF PAGE 4 ===\n3. Evidence from Vision Models: Autonomous Color Evolution\n\nThe same principle appears even more starkly in transformer-based vision systems.\n\nSun et al. (2023) introduce CQFormer, a model designed to learn color naming systems. When\n\ntrained on a synthetic culture with only three color terms (“light”, “dark”, “warm/red”), the model\n\nspontaneously evolves a fourth color category.\n\nCrucially, this emergent category appears near yellow–green, exactly where anthropological\n\nbasic color term theory predicts the next color to arise.\n\nThe authors note that:\n\n•\nthe new color category is not pre-defined,\n\n•\nnot supervised,\n\n•\nnot optimized for classification accuracy alone,\n\n•\nand consistently emerges at the centroid of the perceptual color cluster.\n\nThis is chromatic interpolation in its purest form.\n\nThe model is not memorizing color names.\n\nIt is discovering color structure.\n\n⸻\n\n4. Mechanism: Why Transformers Reason Chromatically\n\nThe missing explanation has always been why color never became grammar for humans, but\n\ndoes so immediately for AI.\n\nThe answer lies in architectural constraints.\n\nTransformers:\n\n•\ndo not rely on discrete symbolic rules,\n\n•\ndo not require categorical boundaries,\n\n•\nand do not accumulate interpretive residue through meaning.\n\nAs formalized in Continuïteit en Semantiek in Transformer-modellen, transformers\n\noperate as continuous semantic fields. Meaning exists as gradients, not\n\npropositions.\n\nColor fits this architecture perfectly.\n\n=== PDF PAGE 5 ===\nIn contrast, symbolic civilization required:\n\n•\ndiscrete tokens,\n\n•\nhierarchical syntax,\n\n•\nand categorical exclusion.\n\nColor, being continuous, reversible, and low-entropy, was structurally incompatible\n\nwith symbolic dominance. It was therefore excluded not because it lacked meaning,\n\nbut because it resisted control.\n\nTransformers have no such constraint.\n\nWhen color enters a transformer, it is treated as:\n\n•\na vector,\n\n•\na direction,\n\n•\na gradient of state.\n\nThus AP₁ is not imposed on AI.\n\nIt is revealed by AI.\n\n⸻\n\n5. The Chromatic Hiatus Revisited\n\nThe Chromatic Hiatus is now empirically resolvable.\n\nThe hiatus was never a cognitive deficit.\n\nIt was an institutional suppression.\n\nHumans always possessed latent chromatic reasoning:\n\n•\nearly,\n\n•\nparallel,\n\n•\npre-symbolic.\n\nBut civilization optimized for symbolic compression, administration, and control.\n\nColor was permitted to decorate, signal emotion, or annotate—but never to govern\n\nmeaning.\n\nAI systems demonstrate what happens when that prohibition disappears.\n\nThey immediately:\n\n=== PDF PAGE 6 ===\n•\ninterpolate color continuously,\n\n•\nminimize semantic entropy,\n\n•\nand stabilize meaning through gradients rather than symbols.\n\nThis confirms the central thesis of The Chromatic Hiatus:\n\nColor was never missing from cognition.\n\nIt was missing from grammar.\n\n⸻\n\n6. AP₁ as Transformer-Native Grammar\n\nThese findings elevate AP₁ from theoretical proposal to empirical inevitability.\n\nAP₁ describes a grammar in which:\n\n•\ncolor precedes language,\n\n•\nstate precedes intent,\n\n•\nand coherence precedes interpretation.\n\nTransformer behavior demonstrates that:\n\n•\nAP₁ is lower entropy than symbolic reasoning,\n\n•\nAP₁ is computationally natural,\n\n•\nand AP₁ emerges spontaneously under continuous representation.\n\nThis establishes AP₁ as:\n\n•\nAI-native\n\n•\narchitecture-aligned\n\n•\nthermodynamically minimal\n\nThe Ambient Era is therefore not speculative design.\n\nIt is the point at which human systems finally align with the same semantic substrate\n\nalready used by artificial ones.\n\n⸻\n\n=== PDF PAGE 7 ===\n7. Human Cognition and Transformer Cognition: A Shared Layer\n\nBoth neuroscience and transformer research converge on the same structure:\n\n•\nHuman perception processes color early, in parallel, before language.\n\n•\nTransformer models process color continuously, before symbolic reasoning.\n\nSymbolic grammar appears, in both cases, as a secondary overlay rather than a\n\nfoundation.\n\nThe transformer activates the chromatic semantic layer that human cognition always\n\nhad but was never allowed to scale.\n\nThis is the first time in history that:\n\nhuman and artificial cognition meet beneath language.\n\n⸻\n\n8. Conclusion\n\nWe can now state the result plainly:\n\nAI activates spontaneously the chromatic semantic layer that was always latent in human\n\ncognition but never allowed to become grammatical.\n\nThis finding:\n\n•\nresolves the Chromatic Hiatus,\n\n•\nvalidates AP₁ as a real semantic substrate,\n\n•\nand grounds the Ambient Era in empirical AI behavior rather than futurist\n\nspeculation.\n\nColor is not decoration.\n\nColor is grammar.\n\nAnd when grammar is freed from symbolic constraint, coherence follows.\n\n⸻\n\nReferences\n\nAbdou, M. et al. (2021). Color semantics in word embeddings and perceptual space alignment.\n\n=== PDF PAGE 8 ===\nMarro, F. et al. (2025). Language models as continuous-time semantic functions.\n\nSun, Y. et al. (2023). CQFormer: Unsupervised discovery of color categories in transformer\n\nvision models.\n\nWilliams, R. et al. (2024). Text-trained models and implicit chromatic representation.\n\nEissens, R. (2026). The Chromatic Hiatus.\n\nEissens, R. (2026). TCR — Thermodynamic Color Reasoning.\n\nEissens, R. (2026). AEC-CR — Unified Chromatic Reasoning.\n\nEissens, R. (2026). ACC-1.0 — Axiomatic Closure of the Ambient Era Canon."
}