List your company

Benchmark Subset Selection on LAM Evaluation Benchmark 40 tasks

0.977Pearson CorrelationCombined Embedding
0.423720.567360.7110.85464Apr 20, 2026

Evaluation results

MethodLinks
Combined EmbeddingSubset Size (n)=2002026.040.977———
Difficulty-basedSubset Size (n)=2002026.040.964———
Combined EmbeddingSubset Size (n)=1002026.040.963———
IRT-basedSubset Size (n)=2002026.040.96———
Random SamplingSubset Size (n)=2002026.040.959———
Anchor PointsSubset Size (n)=2002026.040.952———
Acoustic EmbeddingSubset Size (n)=2002026.040.943———
Random-Sampling-LearnSubset Size (n)=2002026.040.94———
Anchor PointsSubset Size (n)=1002026.040.94———
Random-Search-LearnSubset Size (n)=2002026.040.937———
Semantic EmbeddingSubset Size (n)=2002026.040.936———
Combined EmbeddingSubset Size (n)=502026.040.934———
Difficulty-basedSubset Size (n)=1002026.040.924———
Semantic EmbeddingSubset Size (n)=1002026.040.921———
IRT-basedSubset Size (n)=1002026.040.919———
Random SamplingSubset Size (n)=1002026.040.916———
Anchor PointsSubset Size (n)=502026.040.907———
Acoustic EmbeddingSubset Size (n)=1002026.040.904———
Random-Search-LearnSubset Size (n)=1002026.040.894———
Random-Sampling-LearnSubset Size (n)=1002026.040.887———
Anchor PointsSubset Size (n)=302026.040.884———
Combined EmbeddingSubset Size (n)=302026.040.878———
Semantic EmbeddingSubset Size (n)=502026.040.877———
Acoustic EmbeddingSubset Size (n)=502026.040.87———
IRT-basedSubset Size (n)=502026.040.864———
Difficulty-basedSubset Size (n)=502026.040.863———
Random SamplingSubset Size (n)=502026.040.856———
Anchor PointsSubset Size (n)=202026.040.856———
Combined EmbeddingSubset Size (n)=202026.040.831———
Difficulty-basedSubset Size (n)=302026.040.811———
Variance-basedSubset Size (n)=2002026.040.804———
Random-Search-LearnSubset Size (n)=502026.040.803———
Anchor PointsSubset Size (n)=102026.040.797———
Random-Sampling-LearnSubset Size (n)=502026.040.791———
Random SamplingSubset Size (n)=302026.040.784———
Semantic EmbeddingSubset Size (n)=302026.040.781———
IRT-basedSubset Size (n)=302026.040.778———
Difficulty-basedSubset Size (n)=202026.040.761———
Variance-basedSubset Size (n)=1002026.040.756———
Acoustic EmbeddingSubset Size (n)=102026.040.736———
Random-Search-LearnSubset Size (n)=302026.040.734———
Random-Sampling-LearnSubset Size (n)=302026.040.719———
Random SamplingSubset Size (n)=202026.040.718———
Variance-basedSubset Size (n)=502026.040.716———
IRT-basedSubset Size (n)=202026.040.698———
Random-Search-LearnSubset Size (n)=202026.040.676———
Variance-basedSubset Size (n)=302026.040.676———
Acoustic EmbeddingSubset Size (n)=302026.040.672———
Random-Sampling-LearnSubset Size (n)=202026.040.656———
Combined EmbeddingSubset Size (n)=102026.040.651———
Variance-basedSubset Size (n)=202026.040.628———
Semantic EmbeddingSubset Size (n)=202026.040.627———
Random-Search-LearnSubset Size (n)=102026.040.619———
Difficulty-basedSubset Size (n)=102026.040.608———
Random SamplingSubset Size (n)=102026.040.559———
Random-Sampling-LearnSubset Size (n)=102026.040.544———
Variance-basedSubset Size (n)=102026.040.525———
IRT-basedSubset Size (n)=102026.040.486———
Semantic EmbeddingSubset Size (n)=102026.040.466———
Acoustic EmbeddingSubset Size (n)=202026.040.445———
Acoustic Embeddingn range=[10, 200]2026.04—0.8592250
Anchor Pointsn range=[10, 200]2026.04—0.92740155
Combined Embeddingn range=[10, 200]2026.04—0.9433267
Difficulty-basedn range=[10, 200]2026.04—0.90271157
IRT-basedn range=[10, 200]2026.04—0.89281156
Random Samplingn range=[10, 200]2026.04—0.89183164
Random-Sampling-Learnn range=[10, 200]2026.04—0.854119300
Random-Search-Learnn range=[10, 200]2026.04—0.86699300
Semantic Embeddingn range=[10, 200]2026.04—0.85660350
Variance-basedn range=[10, 200]2026.04—0.742——