HUGIML interactive demos

Explore HUGIML through interactive benchmark, causal-analysis, and governance dashboards, guided browser-playable video demonstrations, and chat-first LLM model-review examples.

HUG-IML IEEE Access 2024 Interactive dashboards Causal analysis LLM examples
Benchmark dashboards
📊
Internal benchmark

Benchmark Analysis Dashboard

100-dataset evaluation, with 50 real-world and 50 synthetic datasets, covering predictive quality, timing, robustness, RPTE behavior, and inspection complexity.

100 datasetsReal and syntheticComplexity
🏆
OpenML-CC18

OpenML-CC18 Benchmark Dashboard

50 completed tasks from the 72-task catalog, evaluated on official outer splits with nested model selection and matched ensemble comparisons.

50 / 72 tasksOfficial splitsNested selection
📈
TabZilla

TabZilla Benchmark Dashboard

Rotating train, validation, and test evaluation on 31 matched datasets with predictive, timing, RPTE, and complexity analysis.

31 datasetsRotating validationMatched models
🧪
PMLBmini

PMLBmini Benchmark Dashboard

All 44 binary-classification datasets evaluated with repeated rotating folds and matched model comparisons.

44 datasetsRepeated foldsBinary classification
🏆
TabArena

TabArena Official Leaderboard

Official-reference Elo and detailed metric analysis with overall, binary, multiclass, default, tuned, and combined views.

Official referenceEloMetric deltas
Scalability

Scalability Dashboard

Measured fit-time and inference-latency scaling against XGBoost and LightGBM across row, feature, and mining-budget sweeps.

Rows to 10MnFeatures to 10,000Latency
Causal Investigation Studio
Causal analysis preview

Causal Investigation Studio Demo

Interactive preview of shared-vocabulary T-HUG treatment-effect analysis: potential outcomes, repeated cross-fitting, DR/AIPW estimates, overlap sensitivity, interpretable causal regions, and matched T-learner comparisons.

T-HUG CATE Cross-fitting DR/AIPW Overlap Regions
Governance Studio
🏛️
Governance preview

Governance Studio Demo

Static preview of the Dash-based Governance Studio: validation evidence, representation audit, HUG pattern inventory, case-level explanations, representation pruning, data quality review, and monitoring signals.

Validation Patterns Pruning Monitoring Data quality Case review
Guided video demos
▶️
Browser-playable demos

HUGIML Guided Video Demonstrations

Short guided walkthroughs covering the HUGIML overview, data-scientist workflow, data-to-decision modeling, interpretability, stakeholder communication, and governance. Videos play directly in the browser with no local setup required.

Overview Workflow Data to decision Interpretability Stakeholder Governance
LLM examples
🤖
LLM interface mock

LLM Chat Workbench

Chat-style workbench mock covering dataset profiling, feature importance, pattern breakdown, model pruning, rebuild comparison, and governance artifact preparation.

Chat workflow Pattern details Feature importance Prune + rebuild
💬
LLM example

Lending Credit-Risk Assistant

Chat-style walkthrough for building, tuning, scoring, and explaining an interpretable lending model — covering risk drivers, pattern summaries, and governance preparation for credit committee review.

Credit risk Q&A flow Decision support Rule explanations
💳
LLM example

Card Default Assistant

Chat-style collections workflow for profiling cardholder risk, comparing rule sets, explaining high-risk accounts, and trimming the model representation for hand-off.

Default risk Collections Governance Pruning