Language Without Propositions: Why Large Language Models
Hallucinate

J 2026

Language Without Propositions: Why Large Language Models Hallucinate

MÁCHA, Jakub

Základní údaje

Originální název

Language Without Propositions: Why Large Language Models Hallucinate

Autoři

MÁCHA, Jakub

Vydání

Philosophies, MDPI, 2026, 2409-9287

Další údaje

Jazyk

angličtina

Typ výsledku

Článek v odborném periodiku

Obor

60300 6.3 Philosophy, Ethics and Religion

Stát vydavatele

Švýcarsko

Utajení

není předmětem státního či obchodního tajemství

Odkazy

URL

Impakt faktor

Impact factor: 0.700 v roce 2024

Označené pro přenos do RIV

Ano

Organizační jednotka

Filozofická fakulta

DOI

https://doi.org/10.3390/philosophies11020042

Klíčová slova anglicky

large language models; artificial intelligence; proposition; fact; truth; hallucination; logical atomism; radical interpretation

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 20. 3. 2026 15:25, prof. Dr. phil. Jakub Mácha, Ph.D.

Anotace

V originále

This paper defends the thesis that LLM hallucinations are best explained as a truth representation problem: Current models lack an internal representation of propositions as truth-bearers, so truth and falsity cannot constrain generation in the way factual discourse requires. It begins by surveying leading explanations—computational limits on self-verification, deficiencies in training data as truth sources, and architectural factors—and argues that they converge on the same underlying representational deficit. Next, it reconstructs the philosophical background of current LLM design, showing how optimization for fluent continuation aligns with coherence-style evaluation and with broadly structuralist, relational semantics, before turning to David Chalmers’s recent attempt to secure propositional interpretability by drawing on Davidson/Lewis-style radical interpretation and by locating propositional content in “middle-layer” structures; it argues that this approach downplays the ubiquity of hallucination and inherits instability from post-training edits. Finally, the paper offers a positive proposal: Atomic propositions should be represented in the basic vector layer, reviving a logical atomist program as a principled route to reducing hallucination.

Návaznosti

MUNI/A/1757/2025, interní kód MU

Název: Nové výzvy v současné estetice

Investor: Masarykova univerzita, Nové výzvy v současné estetice

Citovat

MÁCHA, Jakub. Language Without Propositions: Why Large Language Models Hallucinate. Philosophies. MDPI, 2026, roč. 42, 11(2), s. 1-24. ISSN 2409-9287. Dostupné z: https://doi.org/10.3390/philosophies11020042.

@article{2561725,
   author = {Mácha, Jakub},
   article_number = {11(2)},
   doi = {https://doi.org/10.3390/philosophies11020042},
   keywords = {large language models; artificial intelligence; proposition; fact; truth; hallucination; logical atomism; radical interpretation},
   language = {eng},
   issn = {2409-9287},
   journal = {Philosophies},
   title = {Language Without Propositions: Why Large Language Models Hallucinate},
   url = {https://www.mdpi.com/2409-9287/11/2/42},
   volume = {42},
   year = {2026}
}

TY  - JOUR
ID  - 2561725
AU  - Mácha, Jakub
PY  - 2026
TI  - Language Without Propositions: Why Large Language Models Hallucinate
JF  - Philosophies
VL  - 42
IS  - 11(2)
SP  - 1-24
EP  - 1-24
PB  - MDPI
SN  - 24099287
KW  - large language models
KW  - artificial intelligence
KW  - proposition
KW  - fact
KW  - truth
KW  - hallucination
KW  - logical atomism
KW  - radical interpretation
UR  - https://www.mdpi.com/2409-9287/11/2/42
N2  - This paper defends the thesis that LLM hallucinations are best explained as a truth representation problem: Current models lack an internal representation of propositions as truth-bearers, so truth and falsity cannot constrain generation in the way factual discourse requires. It begins by surveying leading explanations—computational limits on self-verification, deficiencies in training data as truth sources, and architectural factors—and argues that they converge on the same underlying representational deficit. Next, it reconstructs the philosophical background of current LLM design, showing how optimization for fluent continuation aligns with coherence-style evaluation and with broadly structuralist, relational semantics, before turning to David Chalmers’s recent attempt to secure propositional interpretability by drawing on Davidson/Lewis-style radical interpretation and by locating propositional content in “middle-layer” structures; it argues that this approach downplays the ubiquity of hallucination and inherits instability from post-training edits. Finally, the paper offers a positive proposal: Atomic propositions should be represented in the basic vector layer, reviving a logical atomist program as a principled route to reducing hallucination.
ER  -

MÁCHA, Jakub. Language Without Propositions: Why Large Language Models Hallucinate. \textit{Philosophies}. MDPI, 2026, roč.~42, 11(2), s.~1-24. ISSN~2409-9287. Dostupné z: https://doi.org/10.3390/philosophies11020042.

Přehled o publikaci