Universal Conceptual Structure in Neural Translation: Probing NLLB-200’s Multilingual Geometry

universal-conceptual-structure-in-neural-translation:-probing-nllb-200’s-multilingual-geometry

Universal Conceptual Structure in Neural Translation: Probing NLLB-200’s Multilingual Geometry

[Submitted on 27 Feb 2026 (v1), last revised 2 Nov 2026 (this version, v2)]

View PDF HTML (experimental)

Abstract:Do multilingual translation models organize meaning around a shared conceptual space, or do they primarily encode language identity and surface form? We probe the encoder geometry of Meta’s 600M-parameter NLLB-200 model across 135 languages and 101 Swadesh concepts in contextualized carrier-frame and decontextualized bare-word regimes against null and confound controls. First, two-way variance partitioning reveals an additive language-shift plus conceptual-manifold geometry: language shifts account for 45.7% of variance on bare words vs. 13.7% in context, and mean-centering improves the between-to-within-concept distance ratio (vs. a 0.99 permutation null) by 2.14x on bare words (1.15 to 2.46; 2.27x across disjoint scripts) and 1.19x in context (2.25 to 2.69), with conceptual separation peaking at Layer 10 (mutual k-NN overlap jumping 4.35x). Second, semantic category explains 25.4% of Swadesh convergence variance ($F(7,93)=4.53$, $p=2.21times 10^{-4}$; partial $F=7.74$ controlling for fertility and length)—over 20x orthography or phonology ($R^2le 0.012$)—while the advantage of Swadesh over non-loanword concrete nouns ($d=0.54$–$0.69$) is mediated by Zipfian subword compactness ($p=7.43times 10^{-7}$). Third, language distances track ASJP NLD and LDND phonetic distances within subfamilies under block-permutation and cross-script controls ($rho_s=0.65$–$0.67$, block $p=10^{-4}$; global bare-word Mantel $rho_s=0.24$, $p=0.001$), and CLICS$^3$ colexified concepts lie closer across string-distinct, within-category ($rho_s=0.32$), and bare-word ($d=1.79$) controls ($d=1.43$, $p=1.55times 10^{-33}$). Finally, basic color terms recover the Berlin and Kay hierarchy ($rho_s=0.91$–$0.93$, $ple 2.18times 10^{-4}$), and semantic offsets exceed distance-matched controls ($p=3.55times 10^{-5}$, $d=1.06$). We release InterpretCognates, an interactive toolkit and reproducible suite.

Submission history

From: Kyle Mathewson [view email]
[v1] Fri, 27 Feb 2026 22:51:01 UTC (999 KB)
[v2] Mon, 2 Nov 2026 20:17:07 UTC (1,134 KB)