Update README.md
Browse files
README.md
CHANGED
|
@@ -19,8 +19,11 @@ tags:
|
|
| 19 |
|
| 20 |
|
| 21 |
|
| 22 |
-
|
|
|
|
| 23 |
## The first 7b model achieving an 8.8 on MT-Bench which is performance at Humanalities, Coding and Writing.
|
|
|
|
|
|
|
| 24 |
<p
|
| 25 |
align="center"
|
| 26 |
<a href="The TOP1 MT-Bench Model">xDAN-AI</a> •
|
|
@@ -36,11 +39,11 @@ tags:
|
|
| 36 |
|
| 37 |
|
| 38 |
|
| 39 |
-
|
| 40 |
| model | turn | score | size
|
| 41 |
|--------------------|------|----------|--------
|
| 42 |
| gpt-4 | 1 | 8.95625 | -
|
| 43 |
-
| xDAN-L1-Chat-RL-v1 | 1 | 8.87500 | 7b
|
| 44 |
| xDAN-L2-Chat-RL-v2 | 1 | 8.78750 | 30b
|
| 45 |
| claude-v1 | 1 | 8.15000 | -
|
| 46 |
| gpt-3.5-turbo | 1 | 8.07500 | 20b
|
|
@@ -55,8 +58,7 @@ tags:
|
|
| 55 |
|--------------------|------|-----------|--------
|
| 56 |
| gpt-4 | 2 | 9.025000 | -
|
| 57 |
| xDAN-L2-Chat-RL-v2 | 1 | 8.087500 | 30b
|
| 58 |
-
|
|
| 59 |
-
| xDAN-L1-Chat-RL-v1 | 2 | 7.825000 | 7b
|
| 60 |
| gpt-3.5-turbo | 2 | 7.812500 | 20b
|
| 61 |
| claude-v1 | 2 | 7.650000 | -
|
| 62 |
| wizardlm-30b | 2 | 6.887500 | 30b
|
|
@@ -68,22 +70,21 @@ tags:
|
|
| 68 |
|--------------------|-----------|--------
|
| 69 |
| gpt-4 | 8.990625 | -
|
| 70 |
| xDAN-L2-Chat-RL-v2 | 8.437500 | 30b
|
| 71 |
-
| xDAN-L1-Chat-RL-v1 | 8.350000 | 7b
|
| 72 |
| gpt-3.5-turbo | 7.943750 | 20b
|
| 73 |
| claude-v1 | 7.900000 | -
|
| 74 |
| vicuna-33b-v1.3 | 7.121875 | 33b
|
| 75 |
| wizardlm-30b | 7.009375 | 30b
|
| 76 |
| Llama-2-70b-chat | 6.856250 | 70b
|
| 77 |
|
| 78 |
-
|
| 79 |
-
|
| 80 |
Prompt Template(Alpaca)
|
| 81 |
-
### Instruction:
|
| 82 |
-
{
|
| 83 |
### Response:
|
| 84 |
|
| 85 |
|
| 86 |
## Created By xDAN-AI at 2023-12-15
|
|
|
|
| 87 |
## Check: https://www.xdan.ai
|
| 88 |
|
| 89 |
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
|
| 22 |
+
<p
|
| 23 |
+
align="center"
|
| 24 |
## The first 7b model achieving an 8.8 on MT-Bench which is performance at Humanalities, Coding and Writing.
|
| 25 |
+
>
|
| 26 |
+
|
| 27 |
<p
|
| 28 |
align="center"
|
| 29 |
<a href="The TOP1 MT-Bench Model">xDAN-AI</a> •
|
|
|
|
| 39 |
|
| 40 |
|
| 41 |
|
| 42 |
+
**########## First turn ##########**
|
| 43 |
| model | turn | score | size
|
| 44 |
|--------------------|------|----------|--------
|
| 45 |
| gpt-4 | 1 | 8.95625 | -
|
| 46 |
+
| **xDAN-L1-Chat-RL-v1** | 1 | 8.87500 | 7b
|
| 47 |
| xDAN-L2-Chat-RL-v2 | 1 | 8.78750 | 30b
|
| 48 |
| claude-v1 | 1 | 8.15000 | -
|
| 49 |
| gpt-3.5-turbo | 1 | 8.07500 | 20b
|
|
|
|
| 58 |
|--------------------|------|-----------|--------
|
| 59 |
| gpt-4 | 2 | 9.025000 | -
|
| 60 |
| xDAN-L2-Chat-RL-v2 | 1 | 8.087500 | 30b
|
| 61 |
+
| **xDAN-L1-Chat-RL-v1** | 2 | 7.825000 | 7b
|
|
|
|
| 62 |
| gpt-3.5-turbo | 2 | 7.812500 | 20b
|
| 63 |
| claude-v1 | 2 | 7.650000 | -
|
| 64 |
| wizardlm-30b | 2 | 6.887500 | 30b
|
|
|
|
| 70 |
|--------------------|-----------|--------
|
| 71 |
| gpt-4 | 8.990625 | -
|
| 72 |
| xDAN-L2-Chat-RL-v2 | 8.437500 | 30b
|
| 73 |
+
| **xDAN-L1-Chat-RL-v1** | 8.350000 | 7b
|
| 74 |
| gpt-3.5-turbo | 7.943750 | 20b
|
| 75 |
| claude-v1 | 7.900000 | -
|
| 76 |
| vicuna-33b-v1.3 | 7.121875 | 33b
|
| 77 |
| wizardlm-30b | 7.009375 | 30b
|
| 78 |
| Llama-2-70b-chat | 6.856250 | 70b
|
| 79 |
|
|
|
|
|
|
|
| 80 |
Prompt Template(Alpaca)
|
| 81 |
+
### Instruction:"You are a helpful assistant named DAN.You are an expert in worldly knowledge, skilled in employing a probing questioning strategy, carefully considering each step before providing answers."
|
| 82 |
+
{Question}
|
| 83 |
### Response:
|
| 84 |
|
| 85 |
|
| 86 |
## Created By xDAN-AI at 2023-12-15
|
| 87 |
+
## Eval by FastChat: https://github.com/lm-sys/FastChat.git
|
| 88 |
## Check: https://www.xdan.ai
|
| 89 |
|
| 90 |
|