Embedded Files
Activity-Based Costing Framework for Total Cost of Ownership Analysis of LLM Services
Activity-Based Costing Framework for Total Cost of Ownership Analysis of LLM Services
The proliferation of Large Language Models (LLMs) requires robust financial planning, yet traditional software cost models cannot capture the unique economics of generative AI. This paper presents a formal Total Cost of Ownership (TCO) framework grounded in Activity-Based Costing (ABC). We deliver a dual-layered accounting architecture: a mathematically tractable, provably convex engine for technical infrastructure costs, alongside a comprehensive, modular taxonomy blueprint for human operational, governance, and organizational cost pools. The framework systematically incorporates LLM-specific cost drivers, including token consumption, Retrieval-Augmented Generation (RAG) operations, and agentic inference steps. To demonstrate practical applicability, the framework is subjected to a multi-faceted empirical validation programme: the technical infrastructure submodel is validated on a 12-month production RAG chatbot deployment, yielding a predictive formula that links high-level business metrics directly to infrastructure expenses with sub-5% forecasting error, and is further exercised on a second institutional deployment through a synthetically augmented dataset of 95,150 user sessions. Complementary cross-case scenario simulations—spanning regulated, high-scale API, and autonomous agentic regimes—together with hybrid local–cloud serving and retrieval-versus-fine-tuning comparisons, illustrate the framework’s structural generalizability across distinct governance, throughput, and multi-step execution regimes. By integrating time-varying vendor prices, stochastic uncertainty, and hybrid deployment extensions, this work provides a rigorous, transparent decision-support tool for the strategic financial management of LLM-based services.
Post Date: 24 August 2026
Semantic rulers: Continuous measurement of abstract concepts in LLM embedding spaces
Semantic rulers: Continuous measurement of abstract concepts in LLM embedding spaces
Across scientific, economic, and design disciplines, evaluating abstract qualitative concepts — such as environmental sustainability or aesthetic density — traditionally relies on subjective human assessment. This inherent subjectivity limits the integration of these complex traits into automated quantitative workflows. To address this gap, this paper proposes a methodology that uses the latent embedding spaces of large language models and multimodal foundation models to construct continuous semantic scoring instruments. We introduce a counterfactual prompt engineering framework to extract concept vectors from text and image representations while controlling for domain-specific variation, thereby creating calibrated semantic scales. Across the examined domains, the results provide empirical evidence that the target concepts exhibit approximately linear geometric structure. We further evaluate the framework against an externally derived, geospatially grounded ordinal reference (NUS Global Streetscapes with Global Human Settlement Layer typologies), observing a substantial association between the semantic scores and physical settlement density without relying solely on AI self-consistency. The primary novelty of this work lies in extending the use of concept vectors from internal model interpretability toward external semantic scoring. To illustrate a bridge between natural language processing and applied mathematics, we integrate these continuous semantic distances into explicit bounded utility functions. The framework therefore provides a transparent mechanism for separating model-derived semantic scores from normative preference mappings, yielding continuous variables that may serve as components in broader decision support systems.
Post Date: 28 September 2026
Report abuse