A cross-substrate benchmark compares chat-capability across 4 substrates and 3 evaluators. The analysis reveals an anti-correlation pattern between V14 strict scoring and chat capability. The dataset was published by dancinlab on Hugging Face with results dated 2026-05-12.
Use Cases
- Compare chat-capability performance based on the cross-substrate benchmark design.
- Analyze the anti-correlation between V14 strict scoring and chat capability mentioned in the description.
- Evaluate model performance across different evaluation modes (V14-strict, V4-lite chat, V5.8).
- Benchmark AI models using the 4 substrates (A, B', B'', E) detailed in the study.
Strengths
- Benchmark includes results from 4 distinct substrates and 3 different evaluators.
- Analysis identifies a specific decoupling pattern (Lesson Q) between evaluation metrics.
- Results are dated 2026-05-12, indicating recent analysis.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- dancinlab on Hugging Face
- Collection Method
- Likely contains performance metrics from a cross-substrate benchmark evaluation.
- Time Range
- Results dated 2026-05-12.
- Freshness
- Last updated 2026-05-12 01:42:38; freshness should be verified.