Skip to main content
CARLA HQ
PAIRWISE BENCHMARKAdaptive Repeated 3-Fold21 Datasets

TabICLv2 vs EXAONE Tabular Benchmark

Empirical head-to-head evaluation of TabICLv2 (Inria SODA) versus EXAONE Tabular (LG AI Research) on identical, precomputed IID or non-IID splits across 21 diverse tabular datasets.

HEAD-TO-HEAD WINS?Direct pairwise dataset matchup wins where the model achieved lower out-of-sample error loss (1 - AUC for binary, Log-Loss for multiclass, RMSE for regression).
TabICLv26
:
13EXAONE Tabular
(2 ties) across 21 benchmarks
TOURNAMENT ELO?Global Bradley-Terry Elo rating computed across all evaluation-split matchups with 1,000 bootstrap resamples. Higher is better (baseline 1000).
TabICLv2932 (±61)
vs
EXAONE Tabular1021 (±74)
Higher is better • 95% Bootstrap CI
MACRO IMPROVABILITY?Normalized average error regret relative to empirical suite ceilings across all 21 datasets. Lower is better (0.0% is optimal ceiling).
TabICLv22.11%
vs
EXAONE Tabular0.85%
Lower is better • Error regret vs ceiling
MEAN SPLIT LATENCY?Average execution latency in seconds per persisted IID or non-IID evaluation split across all benchmarks. Lower is faster.
TabICLv24.26s
vs
EXAONE Tabular158.67s
Lower is faster • adaptive split protocol
ARCHITECTURAL BLUEPRINTS

Model Specifications & Design Trade-offs

Comparing transformer architecture, Bayesian priors, sequence context windows, and browser execution runtime.

Inria SODA

TabICLv2

GitHub / Paper ↗

Inria SODA's state-of-the-art tabular foundation model engineered for instantaneous in-context learning. TabICLv2 runs zero-shot inference without iterative gradient descent and is fully optimized for client-side WebGPU execution in Google Sheets via Carla HQ.

ArchitectureIn-Context Tabular Transformer with Prefix Attention & KV Caching
Prior / PretrainingSynthetic Copula-based Quantile Priors & Prior.py DAGs
Context Window1,000+ context rows with zero fine-tuning required
Runtime PlatformNative WebGPU Browser Runtime (Carla Engine / ONNX Runtime WASM)
LicenseOpen Source (BSD-3-Clause)
LG AI Research

EXAONE Tabular

GitHub / Paper ↗

Large-scale tabular foundation model by LG AI Research designed for high-capacity tabular representation learning, handling mixed categorical and continuous variables across tabular domains.

ArchitectureMulti-Task Columnar Autoregressive Tabular Transformer
Prior / PretrainingHybrid pre-training on synthetic datasets and diverse real-world tabular corpuses
Context WindowTokenized columnar context with sequence truncation
Runtime PlatformServer-side only (Requires Python runtime / PyTorch / CUDA)
LicenseOpen Source (Apache 2.0)
DEEP DIVE REPORTON-DEVICE RUNTIME ANALYSIS

TabICLv2 vs TabPFN-3 vs TabFM vs EXAONE: Which Model Can Run in the Browser?

Evaluating model weight size (MB vs GB), VRAM requirements, Python server infrastructure overhead, licensing terms, and WebGPU client-side execution compatibility.

HEAD-TO-HEAD BREAKDOWN

21-Dataset Metric Evaluation Matrix

Empirical performance on deterministic, application-appropriate IID or non-IID splits: 1 - AUC for binary classification, Log-Loss for multiclass, and RMSE for regression (all metrics: lower is better).

DatasetDomainTaskRows
Splits?Repeats × folds. IID datasets use three folds with size-aware repeats, and the same row indices are used for every model.
Metric?Primary evaluation error loss: 1 - AUC for binary classification, Log-Loss for multiclass, RMSE for regression. All metrics: lower is better.
TabICLv2
Loss?TabICLv2 primary out-of-sample error loss (lower is better).
EXAONE Tabular
Loss?EXAONE Tabular primary out-of-sample error loss (lower is better).
Margin (Δ)?Error loss difference. Negative (green) indicates TabICLv2 lead; positive (red) indicates EXAONE Tabular lead.
Winner?Model achieving lower out-of-sample error loss on this benchmark split.
TabICLv2
Latency?Mean inference runtime in seconds per evaluation split for TabICLv2.
EXAONE Tabular
Latency?Mean inference runtime in seconds per evaluation split for EXAONE Tabular.
Abalone ↗Agriculture, Forestry & FishingREG4,1773×3RMSE2.032.03+0.00EXAONE Tabular1.07s13.78s
Adult ↗Economics & Public PolicyBIN48,8423×31 - AUC0.07930.0680+0.0113EXAONE Tabular23.84s859.66s
Airfoil Self Noise ↗UCIREG1,50310×3RMSE1.121.16-0.04TabICLv20.65s2.34s
Amazon Employee Access ↗Information Technology & Enterprise SecurityBIN32,7693×31 - AUC0.14830.1232+0.0251EXAONE Tabular11.93s203.54s
Bank Marketing ↗Finance & BankingBIN45,2113×31 - AUC0.19700.1876+0.0094EXAONE Tabular21.49s749.39s
Blood Transfusion Service Center ↗Healthcare & BiomedicineBIN74820×31 - AUC0.24460.2480-0.0035TabICLv20.61s0.24s
Breast W ↗Healthcare & Life SciencesBIN69920×31 - AUC0.00520.0054-0.0002TabICLv20.67s0.30s
Car ↗Automotive & Fleet ManagementMULTI1,72810×3Log-Loss0.02310.0346-0.0115TabICLv20.74s0.56s
Compas Two Years ↗Legal & Public SafetyBIN5,2783×31 - AUC0.27010.2612+0.0089EXAONE Tabular1.73s4.33s
Credit G ↗Finance & BankingBIN1,00010×31 - AUC0.20160.1964+0.0053EXAONE Tabular0.89s0.69s
Diabetes ↗Healthcare & BiomedicineBIN76810×31 - AUC0.16320.1638-0.0005TabICLv20.65s0.30s
Employee Salaries ↗Human Resources & Workforce AnalyticsREG9,2283×3RMSE8,1267,427+699EXAONE Tabular2.13s77.76s
Fitness Club ↗Fitness, Sports & RecreationBIN1,50010×31 - AUC0.17950.1797-0.0002TabICLv20.70s0.49s
House Sales ↗Real Estate & Property ValuationREG21,6133×3RMSE111,52499,520+12,004EXAONE Tabular7.52s931.00s
Houses ↗Real Estate & Urban PlanningREG20,6403×3RMSE40,97139,971+1,000EXAONE Tabular5.04s459.14s
Monks Problems 2 ↗Data Science & Artificial IntelligenceBIN60120×31 - AUC0.00000.0000±0.00Tie0.60s0.24s
Phoneme ↗Speech Processing & Acoustic EngineeringBIN5,4043×31 - AUC0.02780.0276+0.0002EXAONE Tabular1.37s2.54s
Spambase ↗Cybersecurity & IT InfrastructureBIN4,6013×31 - AUC0.00750.0076±0.00Tie3.66s12.63s
Telco Customer Churn ↗Telecommunications & Subscription ServicesBIN7,0433×31 - AUC0.14880.1484+0.0005EXAONE Tabular2.63s11.95s
Titanic ↗Maritime Safety & Actuarial Risk AnalysisBIN1,30910×31 - AUC0.12290.1146+0.0083EXAONE Tabular0.78s0.58s
Vehicle ↗Automotive Engineering & Computer VisionMULTI84610×3Log-Loss0.24010.2393+0.0008EXAONE Tabular0.82s0.53s
ZERO-SERVER CLIENT-SIDE EXECUTION

Evaluate Foundation Models In Your Spreadsheets

Carla HQ brings TabICLv2 foundation models directly into Google Sheets via WebGPU. Run zero-trust predictions on private data without API keys or cloud pipelines.

FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: TabICLv2 vs EXAONE Tabular

Common architectural and benchmarking questions comparing TabICLv2 and EXAONE Tabular.

How does TabICLv2 compare to EXAONE Tabular on tabular benchmarks?

Across the Carla benchmark suite, TabICLv2 and EXAONE Tabular are evaluated head-to-head on identical, precomputed IID or non-IID splits using standardized error losses (1 - AUC for binary classification, Log-Loss for multiclass, and RMSE for regression). IID datasets use adaptive repeated 3-fold evaluation to reduce variance. TabICLv2 enables 100% private in-browser WebGPU execution, whereas server-side models like EXAONE Tabular require heavy PyTorch infrastructure and cloud GPU backend servers.

Can TabICLv2 or EXAONE Tabular run in the browser without server infrastructure?

Only TabICLv2 (via Carla HQ and nanotabicl ONNX Runtime WASM/WebGPU) is engineered to execute entirely client-side inside web browsers like Google Chrome. EXAONE Tabular requires full Python and PyTorch server infrastructure with high memory allocations.

What are the primary architectural differences between TabICLv2 and EXAONE Tabular?

TabICLv2 uses in-context tabular transformer with prefix attention & kv caching, whereas EXAONE Tabular leverages multi-task columnar autoregressive tabular transformer. TabICLv2 specifically utilizes prefix attention with KV cache reuse for real-time tabular evaluation in spreadsheets.