v1.0.4

open-mmlab/Amphionv1.0.4Jul 17, 2026by virattt

AI Summary

Enhances evaluation capabilities with rubric-based scoring and contradiction detection while improving reproducibility and reporting metrics.

Key Highlights

Rubric-based partial-credit scoring and contradiction detection
Reproducible sampling and configurable models
Improved eval reporting with metadata, latency, and failure metrics
Fixes for inaccurate finance evaluation answers

New Features

Rubric-based partial-credit scoring and contradiction detection
Reproducible sampling and configurable models
Improved eval reporting with metadata, latency, and failure metrics
Fix inaccurate finance evaluation answers and rubric criteria

Full Release Notes

### Changes

- Add rubric-based partial-credit scoring and contradiction detection
- Add reproducible sampling, quick runs, configurable models, concurrency, and timeouts
- Improve eval reporting with metadata, latency, and failure metrics

### Fixes

- Fix inaccurate finance evaluation answers and rubric criteria