BTZSC: A Multifaceted Term in Science

• BTZSC is a multifaceted term that denotes diverse constructs such as a zero-shot text classification benchmark, holographic structure constants, geometric compactification programs, and a specialized ceramic composition.
• In NLP, BTZSC standardizes genuine zero-shot evaluation across model families, using controlled protocols that assess semantic matching without supervised tuning.
• Across disciplines, BTZSC reflects the importance of contextual interpretation—from flat-space holography and Lorentzian geometry to material science—necessitating precise bibliographic disambiguation.

BTZSC is a context-dependent acronym or shorthand rather than a single standardized term. In recent arXiv literature it denotes, among other things, a zero-shot text classification benchmark, BTZ-related structure constants in BMS-invariant field theory, a program of BTZ singular compactification in $2+1$-dimensional Lorentzian geometry, smooth complex BTZ saddles for supersymmetric indices, and a medium-entropy BaTiO$_3$-based ceramic composition. In several cases it is the formal name of the object under study; in others it is an interpretive or editorial abbreviation introduced to organize BTZ-adjacent constructions. The meaning of BTZSC is therefore entirely discipline-specific (Aarab, 12 Mar 2026, Bagchi et al., 2023, Khardazi et al., 17 Aug 2025).

1. Terminological landscape

The principal uses of BTZSC in the cited literature are heterogeneous.

BTZSC usage	Domain	Status in source
Benchmark for Textual Zero-Shot Classification	NLP / IR	formal benchmark title (Aarab, 12 Mar 2026)
BTZ-related structure constants	BMSFT / holography	explicit shorthand in technical overview (Bagchi et al., 2023)
BTZ singular compactification program	Lorentzian geometry	interpretive label; not an explicit acronym in the source paper (Brunswic, 2016)
BTZ smooth complex saddles	supersymmetric AdS$_3\times S^3$ gravity	interpretive label summarizing complex BPS saddles (Larsen et al., 27 May 2026)
$0.4$BCZT–$0.6$BSTSn medium-entropy ceramic	ferroelectrics	compositional shorthand (Khardazi et al., 17 Aug 2025)
Band Touchings in the Brillouin Zone: Symmetry Classification	topological band theory	editorial acronym unrelated to BTZ black-hole physics (Wu et al., 2021)

This distribution matters because the same string of letters spans unrelated methodological cultures. In NLP, BTZSC names a controlled benchmark. In holography and Lorentzian geometry, it abbreviates constructions anchored in the Bañados-Teitelboim-Zanelli black hole or its flat-space limits. In materials science, it is simply a compositional shorthand for a lead-free ceramic. The 2016 paper on BTZ extensions is particularly explicit that BTZSC is not an acronym used in the paper itself, but an interpretive label for a compactification program (Brunswic, 2016).

A common source of confusion is therefore assuming that BTZSC has a unique canonical expansion. The literature surveyed here does not support that assumption. A plausible implication is that bibliographic and terminological disambiguation is essential whenever the acronym appears outside its immediate field.

2. BTZSC in zero-shot text classification

In NLP, BTZSC expands to Benchmark for Textual Zero-Shot Classification. It is defined as a unified, controlled evaluation suite for genuine zero-shot text classification across four model families: NLI cross-encoders, embedding models, rerankers, and instruction-tuned LLMs. Its protocol excludes any use of labeled examples from the 22 tasks for training, hyperparameter tuning, or model selection; all tasks are English-only, single-label classification, and all models receive the same verbalized label descriptions (Aarab, 12 Mar 2026).

The benchmark comprises 22 public datasets spanning sentiment, topic, intent, and emotion classification, with diversity in domains, class cardinalities, and document lengths. Evaluation uses each dataset’s official test split where available, otherwise the dev or validation split. BTZSC is designed partly in response to the fact that benchmarks such as MTEB often assess classification through supervised probes on top of frozen embeddings, whereas BTZSC evaluates models purely by semantic matching between input texts and natural-language label descriptions.

The methodological design is deliberately family-agnostic. NLI cross-encoders recast classification as textual entailment and score each text-label pair via entailment logits, using for three-way heads the log-odds

$$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$

Embedding models encode texts and label verbalizers into fixed-size vectors and classify by cosine similarity,

$$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$

Rerankers score text-label pairs directly, while decoder-only generative rerankers such as Qwen3-Reranker compute a forced binary decision probability

$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$

Instruction-tuned LLMs are evaluated in a multiple-choice next-token regime,

$$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$

The primary leaderboard metric is macro-averaged $F_1$,

$_3$0

with accuracy, macro-precision, and macro-recall also reported. The main empirical result is that Qwen3-Reranker-8B achieves a new state of the art with macro $_3$1 and average accuracy $_3$2. Strong embedding models narrow the gap: GTE-large-en-v1.5 reaches $_3$3 $_3$4 with accuracy $_3$5, while 8–12B instruction-tuned LLMs reach up to $_3$6 $_3$7, especially strong on topic classification. NLI cross-encoders remain competitive but plateau, with the best custom DeBERTa-v3-large model reaching $_3$8 $_3$9. Sentiment is relatively easy, emotion is consistently the hardest family, and embedding models often occupy the best accuracy-latency frontier.

BTZSC’s significance in NLP is methodological rather than architectural. It standardizes a genuine zero-shot protocol across model classes that are usually evaluated under incomparable conditions. Its stated limitations are also structural: public datasets may overlap with pretraining corpora, multilingual coverage is absent, and multi-label classification is outside the initial release.

3. BTZSC as BTZ-related structure constants in BMSFT

In a holographic setting, BTZSC denotes BTZ-related structure constants arising from modular covariance of torus two-point functions in two-dimensional BMS-invariant field theories. The underlying motivation is that $_3\times S^3$0d BMSFTs are putative holographic duals of Einstein gravity in $_3\times S^3$1d asymptotically flat spacetimes, and that their torus modular structure can be used to extract averaged off-diagonal three-point coefficients (Bagchi et al., 2023).

The relevant symmetry algebra is the centrally extended $_3\times S^3$2 algebra with generators $_3\times S^3$3 and $_3\times S^3$4. In $_3\times S^3$5d Einstein gravity one has $_3\times S^3$6 and $_3\times S^3$7. The BMS torus partition function depends on modular parameters $_3\times S^3$8, and the BMS $_3\times S^3$9-transformation acts as

$0.4$0

Using modular covariance of the torus two-point function, the paper derives heavy-background expressions for off-diagonal structure constants in two tractable regimes: temporally separated probes with equal angular momentum, and spatially separated probes with equal mass.

The asymptotic density of states is controlled by the BMS-Cardy formula

$0.4$1

which for Einstein gravity reduces to

$0.4$2

Both BTZSC coefficients exhibit exponential suppression governed by the BMS-Cardy entropy, which the paper identifies as an ETH-like signature. The suppression is not exactly $0.4$3, but contains a BMS-specific correction,

$0.4$4

This is the paper’s “interesting change” relative to the $0.4$5d CFT case.

The analytic structure is central. In the temporal channel, the structure constant has a simple pole

$0.4$6

which the paper interprets as encoding perturbations of cosmological horizons in flat-space cosmologies. In the spatial channel, gamma-function poles generate a discrete spectrum,

$0.4$7

These poles are the flat-space remnants of BTZ quasinormal modes. A notable result is that the leading and subleading pieces of the flat limit of BTZ quasinormal frequencies are captured by two different BMSFT off-diagonal structure constants, temporal and spatial respectively.

Within this usage, BTZSC is therefore neither a benchmark nor a geometric spacetime object. It is a set of averaged heavy-heavy-light coefficients whose modular, ETH-like, and pole structures form a precise bridge between BTZ quasinormal-mode data and flat-space holography.

4. BTZSC in geometric compactification and smooth complex saddles

One geometric meaning of BTZSC is the BTZ singular compactification program for globally hyperbolic singular flat spacetimes. Here the basic operation is the controlled addition of extreme BTZ singular lines—parabolic holonomy with lightlike axis—to $0.4$8-dimensional flat Lorentzian spacetimes. The local extreme BTZ model is

$0.4$9

with singular line $0.6$0 lightlike. A BTZ-extension is an injective $0.6$1-morphism whose complement is a union of BTZ lines. The main structural results are that every globally hyperbolic $0.6$2-manifold admits a maximal BTZ-extension unique up to isometry, and that BTZ-extensions preserve Cauchy-maximality and Cauchy-completeness under the hypotheses of Theorem 4.1 (Brunswic, 2016).

The construction is tightly tied to holonomy. Massive particles correspond to elliptic elements with timelike axis, while extreme BTZ lines correspond to parabolic elements fixing a lightlike line. This establishes a Teichmüller–Lorentzian dictionary in which cusps in hyperbolic geometry correspond to BTZ lines and conical points correspond to massive particles. The central point is not black-hole thermodynamics but causal completion: BTZ lines compactify null ends while preserving global hyperbolicity.

A different gravitational use defines BTZSC as BTZ smooth complex saddles in $0.6$3 BPS gravity. In that setting the naïve Euclidean continuation of the Lorentzian BPS black hole fails as the gravitational representative of the supersymmetric index because the contractible Euclidean thermal circle induces antiperiodic fermion boundary conditions and no smooth finite-$0.6$4 cap. The correct saddles are smooth complex finite-temperature BTZ$0.6$5 geometries with complexified angular and $0.6$6-charge chemical potentials (Larsen et al., 27 May 2026).

The canonical $0.6$7d ansatz exhibits a BTZ base fibered with $0.6$8 directions shifted by $0.6$9 and $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$0. The supersymmetric index imposes the holonomy relation

$$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$1

These are simultaneously smoothness conditions at the bolt and global Killing-spinor conditions. The resulting indicial free energy is

$$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$2

The paper derives the same BTZ$$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$3 family in two ways: from $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$4d STU two-center BPS solutions with complex dipoles, and from $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$5d black strings after imposing the BPS relation among thermodynamic potentials.

These two meanings of BTZSC share a BTZ label but not a common mathematical object. One concerns singular line extensions in flat $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$6-gravity; the other concerns smooth complex Euclidean saddles for the supersymmetric index.

5. Derived BTZSC shorthand in adjacent BTZ programs

Several additional uses of BTZSC are best understood as expository shorthand for BTZ-centered constructions rather than stable field-wide nomenclature. In classical string theory, one such use is BTZ String Solutions Construction/Catalog, referring to the dressing-method generation of string solutions in the BTZ background. Starting from timelike or spacelike geodesic seeds in the $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$7 principal chiral model, the dressing construction yields open strings pinned on the boundary, minimal surfaces that can penetrate the horizon, and two BTZ giant-gluon embeddings. For dressed timelike geodesics, the regularized dispersion relation is giant-magnon-like,

$$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$8

while one giant-gluon embedding yields a logarithmic relation $$s = l_{\mathrm{ent}} - \log\!\big(\exp(l_{\mathrm{neut}}) + \exp(l_{\mathrm{contra}})\big).$$9 (David et al., 2012).

In worldsheet BTZ CFT, the paper on the Lorentzian BTZ spectrum argues that the BTZ spectrum is plausibly that of $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$0, rewritten in a hyperbolic basis adapted to the BTZ orbifold. The analysis quantizes timelike, spacelike, and null BTZ geodesics, identifies discrete and principal continuous series, and rewrites the modular-invariant $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$1 partition function so that BTZ primaries and descendants are visible in the hyperbolic basis (Nippanikar et al., 2021).

In differential geometry, the nonrotating BTZ spacetime admits global algebraic embeddings into flat spaces with two extra dimensions. The Lorentzian solution embeds in $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$2, the Euclidean solution in $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$3, and in both cases the image is the intersection of two quadric hypersurfaces. The same paper also shows that nonrotating BTZ is of embedding class one relative to $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$4 or $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$5 (Willison, 2010).

In braneworld gravity, BTZ-like black holes appear on codimension-2 $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$6-branes in five-dimensional gravity with a Gauss–Bonnet term and induced gravity. The effective boundary equations are three-dimensional Einstein equations with a deficit-angle-dependent Planck mass

$$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$7

and the BTZ geometry extends into the bulk as a regular horizon string-like solution under the fine-tuned branch with $$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$8 (Cuadros-Melgar et al., 2010).

A quantum-mechanical reinterpretation appears in the Chern–Simons matrix model for BTZ entropy. There the quadratic invariant

$$s(x,y)=\frac{\langle e(x),e(y)\rangle}{\|e(x)\|\,\|e(y)\|}.$$9

has oscillator spectrum

$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$0

and its degeneracy grows like a partition number. Identifying $P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$1 with $P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$2 reproduces the BTZ entropy formula in the large-$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$3 limit (Chaney et al., 2013).

A further near-horizon usage appears in the study of BPS fivebrane stars. That work constructs a large class of NS5–F1–P BPS solutions with arbitrary chiral profiles. Within that space lies an ensemble of ultracompact, horizonless objects whose exteriors mimic extremal BTZ black holes and develop deep $P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$4 throats. The brane effective action exhibits a collective breathing mode analogous to the Schwarzian mode of $P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$5 black holes, and the approach to the black-hole phase is tied to light D-brane excitations and a fivebrane deconfinement transition (Martinec et al., 9 Dec 2025).

These examples show a recurrent pattern: BTZSC often functions as a compact label for a family of BTZ-related constructions rather than a formally standardized term.

6. BTZSC as a medium-entropy ferroelectric ceramic

In materials science, BTZSC is a compositional shorthand for a BaTiO$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$6-based medium-entropy ceramic,

$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$7

i.e. $P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$8BCZT–$P_{\mathrm{yes}}=\frac{e^{u_{T_{\mathrm{yes}}}}}{e^{u_{T_{\mathrm{yes}}}}+e^{u_{T_{\mathrm{no}}}}.$9BSTSn. The overall A-site occupancy is $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$0 and the B-site occupancy is $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$1, increasing configurational disorder on both sublattices relative to the parent phases (Khardazi et al., 17 Aug 2025).

The ceramic is prepared by a sol–gel route. BCZT and BSTSn are synthesized separately, dried, calcined at $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$2 for $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$3 h, then mixed in mole ratio $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$4, wet-milled with absolute ethanol, dried, pressed into pellets of thickness $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$5 mm, and sintered at $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$6 for $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$7 h. The density used in electrocaloric calculations is $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$8. SEM shows dense ceramics with non-uniform grain-size distribution; the average grain size is $$P(y_k\mid x)=\frac{\exp(v_{T_{a_k}})}{\sum_{j=1}^{K}\exp(v_{T_{a_j}})}, \qquad \hat y=\arg\max_k P(y_k\mid x).$$9 in BTZSC, compared with $F_1$0 in BCZT and $F_1$1 in BSTSn.

Room-temperature XRD shows a phase-pure perovskite without secondary phases, indicating complete solid solubility. Rietveld refinement gives coexistence of tetragonal $F_1$2 and orthorhombic $F_1$3 phases in BTZSC. The tetragonality of the $F_1$4 phase increases relative to BCZT, with refined $F_1$5 in BTZSC versus $F_1$6 in BCZT. Dielectrically, BTZSC has $F_1$7 at $F_1$8 kHz around $F_1$9, with $_3$00. Modified Curie–Weiss analysis places the diffuseness degree $_3$01 between $_3$02 and $_3$03, indicating a diffuse transition with relaxor-like features.

Energy-storage metrics are extracted from the upper branch of the $_3$04–$_3$05 loops,

$_3$06

At $_3$07 and $_3$08 Hz, BTZSC reaches $_3$09 at $_3$10. The efficiency is reported as $_3$11 in the abstract, but the detailed analysis and Table 3 give $_3$12 at $_3$13; this discrepancy is one of the few explicit internal numerical tensions in the paper. Temperature stability is strong: $_3$14 varies by less than $_3$15 between $_3$16 and $_3$17, specifically $_3$18–$_3$19.

The electrocaloric effect is evaluated indirectly through the Maxwell relation,

$_3$20

For BTZSC, the peak electrocaloric temperature change is $_3$21 at $_3$22, with responsivity $_3$23. Both $_3$24 and $_3$25 exceed those of the parent BCZT and BSTSn ceramics under the reported conditions.

The mechanistic interpretation is framed in terms of medium entropy. Configurational entropy on each sublattice is written

$_3$26

giving $_3$27 and $_3$28 for BTZSC. The paper attributes the improved performance to compositional disorder, lattice distortion, multiphase coexistence, increased tetragonality, smaller grains, reduced $_3$29, and easier domain switching. A plausible implication is that BTZSC occupies a design space in which moderate field operation, diffuse transition behavior, and lead-free composition are jointly optimized, although breakdown strength, leakage current, cycling endurance, and long-term reliability were not assessed.

The ceramic usage illustrates the strongest possible departure from BTZ black-hole terminology: here BTZSC is purely a materials shorthand, with no connection to BMSFT, $_3$30, or BTZ spacetime.