Request a Discussion

Benchmarks

Benchmarks for Open Metal AI

Reproducible leaderboard tasks, each with a defined dataset, evaluation metric, baseline model, and submission rules. They give everyone a shared basis for comparing methods on metal-transformation problems.

10 results

Sort

AluminiumChemistry prediction

Aluminium Melt Chemistry Prediction

Predict resulting melt chemistry from charge, additions, and furnace conditions. Lower mean MAE across elements is better.

MetricMean MAE (wt%)

BaselineAluminium Melt Chemistry Predictor

#1 matereal0.031

Rare earthsRecovery prediction

Rare Earth Recovery Prediction

Predict element recovery across a solvent-extraction separation train from stage conditions and feed assay. Lower MAE is better.

MetricMAE (recovery %)

BaselineRare Earth Recovery Predictor

#1 k.almeida1.42

CopperYield / loss prediction

Copper Slag Loss Prediction

Predict copper loss to slag from furnace conditions, matte grade, and slag chemistry. Lower MAE is better.

MetricMAE (Cu-in-slag %)

BaselineCopper Slag Loss Predictor

#1 l.zhang0.052

LithiumImpurity prediction

Lithium Impurity Risk Prediction

Predict impurity breakthrough to battery grade from feed profile and purification conditions. Higher ROC-AUC is better.

BaselineLithium Impurity Risk Model

#1 p.nakamura0.921

Metal powdersQualification classification

Metal Powder Qualification Classification

Classify whether a powder batch qualifies from morphology, chemistry, flowability, and oxygen content. Higher F1 is better.

MetricF1 (qualified)

BaselineMetal Powder Qualification Classifier

#1 p.nakamura0.912

Rare earthsPurity classification

Rare Earth Purity Classification

Classify product purity band from stage chemistry and impurity indicators. Higher macro-F1 is better.

BaselineRare Earth Purity Classifier

#1 h.novak0.905

Scrap & recycled metalsContamination prediction

Scrap Blend Quality Prediction

Predict resulting quality and impurity risk from scrap blend ratios and stream assays. Higher F1 is better.

MetricF1 (contamination)

BaselineSecondary Metals Contamination Risk Model

#1 r.okafor0.861

Nickel & cobaltSelectivity prediction

Nickel-Cobalt Selectivity Prediction

Predict Ni/Co selectivity across a solvent-extraction separation. Lower MAE is better.

MetricMAE (selectivity)

BaselineNickel-Cobalt Separation Predictor

#1 k.almeida0.81

ZincRecovery prediction

Zinc Leach Recovery Prediction

Predict zinc recovery through leaching and purification. Lower MAE is better.

MetricMAE (recovery %)

BaselineZinc Leach Recovery Predictor

#1 matereal1.50

FerroalloysChemistry prediction

Ferroalloy Composition Prediction

Predict tapped alloy composition from furnace charge and operating conditions. Lower MAE is better.

MetricMAE (wt%)

BaselineFerroalloy Composition Predictor

#1 matereal0.30

Propose a benchmark

Propose a benchmark or submit to a leaderboard

Benchmarks are open and community-governed. Propose a new task, contribute a submission, or discuss metric design with other builders.

Join the Community Explore Open Models