scores/feature.txt

================================================================================
EVALUATION RESULTS
================================================================================

📊 Accuracy Metrics:
  Exact Match Accuracy:  50.17% (63805/135256)
  VQA Accuracy:          15.72%

📊 ANLS Metrics:
  Average ANLS (τ=0.5):  50.18%
  ANLS Std Dev:          48.96%

📊 Additional Statistics:
  Total samples:         135256
  Avg prediction length: 1.13 words
  Avg GT length:         1.10 words

================================================================================
SAMPLE PREDICTIONS
================================================================================

🏆 Best Predictions (Highest ANLS):
--------------------------------------------------------------------------------

Ground Truth: tusks
Prediction:   tusks
ANLS:         1.0000
Exact Match:  ✓

Ground Truth: seagull
Prediction:   seagull
ANLS:         1.0000
Exact Match:  ✓

Ground Truth: bedroom
Prediction:   bedroom
ANLS:         1.0000
Exact Match:  ✓

Ground Truth: cake
Prediction:   cake
ANLS:         1.0000
Exact Match:  ✓

Ground Truth: short
Prediction:   short
ANLS:         1.0000
Exact Match:  ✓

================================================================================
⚠️  Worst Predictions (Lowest ANLS):
--------------------------------------------------------------------------------

Ground Truth: mirror
Prediction:   car
ANLS:         0.0000
Exact Match:  ✗

Ground Truth: towel
Prediction:   toy
ANLS:         0.0000
Exact Match:  ✗

Ground Truth: book
Prediction:   camera
ANLS:         0.0000
Exact Match:  ✗

Ground Truth: usa
Prediction:   england
ANLS:         0.0000
Exact Match:  ✗

Ground Truth: red and yellow
Prediction:   green
ANLS:         0.0000
Exact Match:  ✗