Update README.md

README.md

@@ -51,6 +51,20 @@ All pre-training is done on the [Cultura-X](https://huggingface.co/datasets/uonl
 ## Tokenizer Details
 We extended the vocabulary of the base llama model from 32,000 tokens to 57,000 tokens by adding up to 25,000 non-overlapping tokens from the new language.
 
+## Evaluation
+
+| | SambaLingo-Thai-Base         | typhoon-7b | bloom-7b1 | xglm-7.5B | mGPT-13B |
+|------------------------------|------------|-----------|-----------|----------|--------|
+| Perplexity (Lower Is Better) | **1.288**      | 1.373     | 1.834     | 1.394    | 1.966  |
+| FLORES en->th (8 shot, CHRF) | **0.433**      | 0.347     | 0.095     | 0.198    | 0.032  |
+| FLORES th->en (8 shot, CHRF) | **0.536**      | 0.465     | 0.138     | 0.431    | 0.016  |
+| FLORES en->th (8 shot, BLEU) | **0.019**      | 0.004     | 0.000     | 0.003    | 0.000  |
+| FLORES th->en (8 shot, BLEU) | **0.247**      | 0.188     | 0.003     | 0.147    | 0.000  |
+| Belebele (3 shot)            | 37.11%     | **52.22%**    | 24.11%    | 22.44%   | 26.89% |
+| SIB-200 (3 shot)             | 62.25%     | **75.49%**    | 23.04%    | 63.73%   | 44.12% |
+| XCOPA (0 shot)               | **61.40%**     | 60.60%    | 55.40%    | 59.40%   | 52.80% |
+| XNLI (0 shot)                | **44.65%**     | 43.01%    | 34.87%    | 43.73%   | 39.24% |
+
 ## Uses
 <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->