---
license: afl-3.0
datasets:
- copenlu/fever_gold_evidence
language:
- en
metrics:
- precision
- recall
- f1
---

```
wandb:                 eval/f1 0.86005
wandb:               eval/loss 0.63145
wandb:                  eval/p 0.85922
wandb:                  eval/r 0.86396
```

**Note**:   
1. `[evidence_text][SEP][claim]`
2. Only trained/validated on instances length <= 512 tokens.