| --- |
| language: |
| - vi |
| - ede |
| tags: |
| - cross-lingual-retrieval |
| - morpheme-tokenizer |
| - colbert |
| - EViRAL |
| --- |
| |
| # ColBERT Transformer + Morpheme Tokenizer v4 — EViRAL |
|
|
| Task: Ede query → Vietnamese passage retrieval |
|
|
| ## Eval Results |
|
|
| | Metric | Validation | Test | |
| |---------|-----------|--------| |
| | nDCG@1 | 0.0004 | 0.0005 | |
| | nDCG@5 | 0.0011 | 0.0013 | |
| | nDCG@10 | 0.0017 | 0.0022 | |
| | MRR@10 | 0.0020 | 0.0023 | |
| | R@50 | 0.0202 | 0.0210 | |
| | R@100 | 0.0345 | 0.0374 | |
|
|
| ## Checkpoints |
| | file | description | |
| |---|---| |
| | mlm.pt | MLM ColBERT encoder weights | |
| | align.pt | cross-lingual aligned encoder | |
| | finetune.pt | contrastive fine-tuned encoder (best val) | |
|
|