Update README.md

Author: edwardjhu

README.md

@@ -17,14 +17,14 @@ We obtain result comparable or superior to full finetuning on the GLUE benchmark
 
 |   | Dataset/Method          | RoBERTa base <br> Fine-tune (125M) |  RoBERTa base <br> LoRA (0.8M) | DeBERTa XXL <br> Fine-tune (1.5B) | DeBERTa XXL <br> LoRA (4.7M)  |
 |---|-------------------------|----------------|--------------------------|-----------------|-----------------|
-|   | MNLI (m-Acc/mm-Acc)     | <b>87.6</b> | [<b>87.5</b>±.3/<b>86.9</b>±.3](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_mnli.bin) |91.7/<b>91.9</b>| [<b>91.9</b>±.1/<b>91.9</b>±.2](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_mnli.bin)       |
-|   | SST2 (Acc)              | 94.8 | [<b>95.1</b>±.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_sst2.bin) | <b>97.2</b>    | [96.9±.2](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_sst2.bin)                    |
-|   | MRPC (Acc)              | <b>90.2</b> | [<b>89.7</b>±.7](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_mrpc.bin) | 92.0           | [<b>92.6</b>±.6](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_mrpc.bin)             |
-|   | CoLA (Matthew's Corr)   | <b>63.6</b> | [<b>63.4</b>±1.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_cola.bin) | <b>72.0</b>    | [<b>72.4</b>±1.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_cola.bin)           |
-|   | QNLI (Acc)              | 92.8 | [<b>93.3</b>±.3](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_qnli.bin) | <b>96.0</b>    | [<b>96.0</b>±.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_qnli.bin)            |
-|   | QQP (Acc)               | <b>91.9</b> | [90.8±.1](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_qqp.bin) | 92.7           | [<b>92.9</b>±.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_qqp.bin)           |
-|   | RTE (Acc)               | 78.7 | [<b>86.6</b>±.7](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_rte.bin) | 93.9           | [<b>94.9</b>±.4](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_rte.bin)           |
-|   | STSB (Pearson/Spearman Corr) | 91.2 | [<b>91.5</b>±.2/<b>91.3</b>±.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa/roberta_base_lora_stsb.bin) |<b>92.9</b>/92.6| [<b>93.0</b>±.2/<b>92.9</b>±.3](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_stsb.bin)      |
+|   | MNLI (m-Acc/mm-Acc)     | <b>87.6</b> | [<b>87.5</b>±.3/<b>86.9</b>±.3](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_mnli.bin) |91.7/<b>91.9</b>| [<b>91.9</b>±.1/<b>91.9</b>±.2](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_mnli.bin)       |
+|   | SST2 (Acc)              | 94.8 | [<b>95.1</b>±.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_sst2.bin) | <b>97.2</b>    | [96.9±.2](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_sst2.bin)                    |
+|   | MRPC (Acc)              | <b>90.2</b> | [<b>89.7</b>±.7](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_mrpc.bin) | 92.0           | [<b>92.6</b>±.6](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_mrpc.bin)             |
+|   | CoLA (Matthew's Corr)   | <b>63.6</b> | [<b>63.4</b>±1.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_cola.bin) | <b>72.0</b>    | [<b>72.4</b>±1.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_cola.bin)           |
+|   | QNLI (Acc)              | 92.8 | [<b>93.3</b>±.3](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_qnli.bin) | <b>96.0</b>    | [<b>96.0</b>±.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_qnli.bin)            |
+|   | QQP (Acc)               | <b>91.9</b> | [90.8±.1](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_qqp.bin) | 92.7           | [<b>92.9</b>±.1](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_qqp.bin)           |
+|   | RTE (Acc)               | 78.7 | [<b>86.6</b>±.7](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_rte.bin) | 93.9           | [<b>94.9</b>±.4](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_rte.bin)           |
+|   | STSB (Pearson/Spearman Corr) | 91.2 | [<b>91.5</b>±.2/<b>91.3</b>±.2](https://github.com/microsoft/LoRA/releases/download/RoBERTa-base/roberta_base_lora_stsb.bin) |<b>92.9</b>/92.6| [<b>93.0</b>±.2/<b>92.9</b>±.3](https://github.com/microsoft/LoRA/releases/download/DeBERTa/deberta_v2_xxlarge_lora_stsb.bin)      |
 
 Fine-tuning numbers taken from [Liu et al. (2019)](https://arxiv.org/abs/1907.11692) and [He et al. (2020)](https://arxiv.org/abs/2006.03654). Please follow the instructions in `NLU/` to reproduce our result.