analiza aiqbench
This research demonstrates that current language models systematically misclassify satisfiable problems involving unresolved circular truth dependencies as contradictions, a phenomenon termed "phantom paradoxes." The study isolates this failure mode to the specific structure of ungrounded cyclic references rather than general self-reference or logical complexity, revealing a fundamental limitation in how models propagate truth values through non-terminating chains.
Phantom Paradoxes: Language Models Treat Unresolved Circular Truth Dependence as Contradiction
arXiv:2604.12128v2 Announce Type: replace Abstract: Do language models struggle with self-reference itself, or with structures where propagating truth values along references cannot settle a value in finitely many steps? Our results support the latter. In our main test, statements refer to one another's truth in a cycle, yet exactly one consistent assignment exists. Local propagation never settles the queried value, but exhaustive search finds it, so INCONSISTENT (no consistent assignment) is always wrong. We call such misreadings of satisfiable problems phantom paradoxes. Each circular problem has a grounded twin with the same answer but no cycle. To rule out logical difficulty and unfamiliar wording, we added constraint twins (equivalent formulas without talk of truth) and perplexity-matched grounded controls. In confirmatory behavioral tests fixed before any confirmation response, four open models from three families answered INCONSISTENT on average 31.0 percentage points more often on circular problems than on grounded twins, 36.4 more than on constraint twins and 37.1 more than on perplexity-matched controls. In each model, validated layer-wise truth directions, fitted on unrelated arithmetic, recovered the queried value at the question's last token for grounded twins but not for circular problems. Constraint twins, likewise unrecovered, drew INCONSISTENT far less often, so circular truth dependence, not the unrecovered value alone, matters. In a small natural-language bank, self-reference settling in finitely many steps drew INCONSISTENT 23.4 points less often than unsettled self-reference with the same classical solution. Pooled tests did not show the predicted late-layer revision or attention spread. Overall, models treat unresolved circular truth dependence as contradiction. We will release all prompts, outputs and hidden states.
Abstract:Do language models struggle with self-reference itself, or with structures where propagating truth values along references cannot settle a value in finitely many steps? Our results support the latter. In our main test, statements refer to one another's truth in a cycle, yet exactly one consistent assignment exists. Local propagation never settles the queried value, but exhaustive search finds it, so INCONSISTENT (no consistent assignment) is always wrong. We call such misreadings of satisfiable problems phantom paradoxes. Each circular problem has a grounded twin with the same answer but no cycle. To rule out logical difficulty and unfamiliar wording, we added constraint twins (equivalent formulas without talk of truth) and perplexity-matched grounded controls. In confirmatory behavioral tests fixed before any confirmation response, four open models from three families answered INCONSISTENT on average 31.0 percentage points more often on circular problems than on grounded twins, 36.4 more than on constraint twins and 37.1 more than on perplexity-matched controls. In each model, validated layer-wise truth directions, fitted on unrelated arithmetic, recovered the queried value at the question's last token for grounded twins but not for circular problems. Constraint twins, likewise unrecovered, drew INCONSISTENT far less often, so circular truth dependence, not the unrecovered value alone, matters. In a small natural-language bank, self-reference settling in finitely many steps drew INCONSISTENT 23.4 points less often than unsettled self-reference with the same classical solution. Pooled tests did not show the predicted late-layer revision or attention spread. Overall, models treat unresolved circular truth dependence as contradiction. We will release all prompts, outputs and hidden states.
v2: major revision with a new title and a confirmatory study. Withdrawn from v1: the attention effective-rank signature and the architecture-invariance claim. Previous title: When Self-Reference Fails to Close: Matrix-Level Dynamics in Large Language Models. 19 pages, 2 figures, 8 tables
Computation and Language (cs.CL)
arXiv:2604.12128 [cs.CL]
(or arXiv:2604.12128v2 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 312 cuvinte
Subiecte conexe
Article: Your Next DSL Author is a Language Model
InfoQ · 16 sept. 2026
Reviewing failure modes in science as we prepare for AI researchers
LessWrong · 8 oct. 2026
Review of Fundamental Uncertainty.
LessWrong · 1 aug. 2026
Readable, Faithful, Used: Three Dissociable Properties of Demographic Identity in a Language Model
arXiv · 20 aug. 2026