modern-pipelines-dissolve-classical-paradigm-taxonomy
IN derived (depth 2)
Created 2026-06-21T10:09:45+00:00 · Reviewed 2026-06-21T15:37:01+00:00
Modern LLM training pipelines dissolve the classical three-paradigm taxonomy — self-supervised pretraining blurs the supervised/unsupervised boundary (its taxonomic status is actively debated), and the full pipeline synthesizes all three paradigms sequentially, suggesting the taxonomy was always a pedagogical convenience rather than a natural partition of learning.
Justifications
SL — Sequential composition of all three paradigms in one pipeline renders the paradigm distinction artificial
Antecedents (all must be IN):
- IN self-supervised-paradigm-boundary-contested — Self-supervised learning occupies a contested taxonomic position — formally a subset of unsupervised learning, yet it has become the dominant pre-training paradigm, creating tension between its classification and its practical centrality.
- IN ml-three-classical-paradigms — The three classical machine learning paradigms are supervised learning (labelled data), unsupervised learning (no labels), and reinforcement learning (reward signal)
Dependents
These beliefs depend on this one:
- OUT decision-transformer-dissolves-rl-sequence-modeling-boundary — Decision Transformer's recasting of reinforcement learning as sequence modeling dissolves the boundary between RL and sequence prediction — combining with the broader dissolution of classical paradigm boundaries (supervised/unsupervised/RL), this represents a specific mechanism by which the Transformer architecture actively drives paradigm taxonomy dissolution rather than merely being affected by it.
- IN deepseek-validates-paradigm-taxonomy-dissolution-in-rl — DeepSeek-R1's achievement of competitive reasoning performance through large-scale RL without supervised fine-tuning further validates paradigm taxonomy dissolution — a traditionally supervised task (reasoning) solved through RL alone, without the supervised intermediate step that the standard LLM pipeline assumes, demonstrating that paradigm boundaries dissolve not only in training pipelines but in task requirements.
- IN ml-conceptual-foundations-doubly-unstable — ML's conceptual foundations are doubly unstable — the classical paradigm taxonomy (supervised/unsupervised/RL) is dissolving as modern pipelines combine all three, while even dominant paradigms like GANs and pretrain-finetune prove empirically fragile and transient — suggesting that ML's organizing categories are descriptive conveniences rather than natural kinds.
- OUT modern-llm-pipeline-sustainable — Modern LLM training pipelines (self-supervised pretraining → instruction tuning → RLHF) are a sustainable methodology — they dissolve classical paradigm boundaries by successfully combining all three ML paradigms, and self-supervised learning provides an effectively unlimited source of training signal.
- IN self-supervised-engine-and-symptom-of-taxonomy-dissolution — Self-supervised learning is both the engine and the symptom of paradigm taxonomy dissolution — it occupies a contested taxonomic position precisely because it is the mechanism through which modern pipelines dissolve the classical supervised/unsupervised boundary, making its own classification impossible under the framework it is dismantling.
- IN transfer-learning-bridges-classical-and-modern-ml — Transfer learning is the conceptual bridge between classical and modern ML — it formalizes classical domain adaptation while simultaneously enabling modern LLM pipelines to dissolve paradigm boundaries, as self-supervised pretraining is precisely transfer learning operating at industrial scale across the supervised/unsupervised divide.