Update README.md
Browse files
README.md
CHANGED
|
@@ -37,6 +37,87 @@ Merges:
|
|
| 37 |
- Tokenizers 0.14.1
|
| 38 |
|
| 39 |
## Evals LM-Evaluation Harness
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
|
| 41 |
`big-refactor` branch:
|
| 42 |
|
|
|
|
| 37 |
- Tokenizers 0.14.1
|
| 38 |
|
| 39 |
## Evals LM-Evaluation Harness
|
| 40 |
+
`mt-bench`:
|
| 41 |
+
```
|
| 42 |
+
Mode: single
|
| 43 |
+
Input file: data/mt_bench/model_judgment/gpt-4_single.jsonl
|
| 44 |
+
|
| 45 |
+
########## First turn ##########
|
| 46 |
+
score
|
| 47 |
+
model turn
|
| 48 |
+
gpt-4 1 8.95625
|
| 49 |
+
claude-v1 1 8.15000
|
| 50 |
+
gpt-3.5-turbo 1 8.07500
|
| 51 |
+
LUNA-SOLARkrautLM-Instruct 1 7.93750
|
| 52 |
+
UNA-SOLAR-10.7B-Instruct-v1.0 1 7.80625
|
| 53 |
+
vicuna-33b-v1.3 1 7.45625
|
| 54 |
+
wizardlm-30b 1 7.13125
|
| 55 |
+
tulu-30b 1 7.01875
|
| 56 |
+
vicuna-13b-v1.3 1 6.81250
|
| 57 |
+
guanaco-65b 1 6.78125
|
| 58 |
+
nous-hermes-13b 1 6.43125
|
| 59 |
+
alpaca-13b 1 4.97500
|
| 60 |
+
rwkv-4-raven-14b 1 4.74375
|
| 61 |
+
llama-13b 1 3.26250
|
| 62 |
+
|
| 63 |
+
########## Second turn ##########
|
| 64 |
+
score
|
| 65 |
+
model turn
|
| 66 |
+
gpt-4 2 9.025000
|
| 67 |
+
gpt-3.5-turbo 2 7.812500
|
| 68 |
+
claude-v1 2 7.650000
|
| 69 |
+
UNA-SOLAR-10.7B-Instruct-v1.0 2 7.237500
|
| 70 |
+
LUNA-SOLARkrautLM-Instruct 2 6.987500
|
| 71 |
+
wizardlm-30b 2 6.887500
|
| 72 |
+
vicuna-33b-v1.3 2 6.787500
|
| 73 |
+
guanaco-65b 2 6.037500
|
| 74 |
+
vicuna-13b-v1.3 2 5.962500
|
| 75 |
+
tulu-30b 2 5.850000
|
| 76 |
+
nous-hermes-13b 2 4.664557
|
| 77 |
+
alpaca-13b 2 4.087500
|
| 78 |
+
rwkv-4-raven-14b 2 3.225000
|
| 79 |
+
llama-13b 2 1.950000
|
| 80 |
+
|
| 81 |
+
########## Average ##########
|
| 82 |
+
score
|
| 83 |
+
model
|
| 84 |
+
gpt-4 8.990625
|
| 85 |
+
gpt-3.5-turbo 7.943750
|
| 86 |
+
claude-instant-v1 7.905660
|
| 87 |
+
claude-v1 7.900000
|
| 88 |
+
UNA-SOLAR-10.7B-Instruct-v1.0 7.521875
|
| 89 |
+
LUNA-SOLARkrautLM-Instruct 7.462500
|
| 90 |
+
vicuna-33b-v1.3 7.121875
|
| 91 |
+
wizardlm-30b 7.009375
|
| 92 |
+
Llama-2-70b-chat 6.856250
|
| 93 |
+
Llama-2-13b-chat 6.650000
|
| 94 |
+
guanaco-33b 6.528125
|
| 95 |
+
tulu-30b 6.434375
|
| 96 |
+
guanaco-65b 6.409375
|
| 97 |
+
oasst-sft-7-llama-30b 6.409375
|
| 98 |
+
palm-2-chat-bison-001 6.400000
|
| 99 |
+
mpt-30b-chat 6.393750
|
| 100 |
+
vicuna-13b-v1.3 6.387500
|
| 101 |
+
wizardlm-13b 6.353125
|
| 102 |
+
Llama-2-7b-chat 6.268750
|
| 103 |
+
vicuna-7b-v1.3 5.996875
|
| 104 |
+
baize-v2-13b 5.750000
|
| 105 |
+
nous-hermes-13b 5.553459
|
| 106 |
+
mpt-7b-chat 5.459119
|
| 107 |
+
gpt4all-13b-snoozy 5.452830
|
| 108 |
+
koala-13b 5.350000
|
| 109 |
+
mpt-30b-instruct 5.218750
|
| 110 |
+
falcon-40b-instruct 5.168750
|
| 111 |
+
h2ogpt-oasst-open-llama-13b 4.625000
|
| 112 |
+
alpaca-13b 4.531250
|
| 113 |
+
chatglm-6b 4.500000
|
| 114 |
+
oasst-sft-4-pythia-12b 4.318750
|
| 115 |
+
rwkv-4-raven-14b 3.984375
|
| 116 |
+
dolly-v2-12b 3.275000
|
| 117 |
+
fastchat-t5-3b 3.040625
|
| 118 |
+
stablelm-tuned-alpha-7b 2.753125
|
| 119 |
+
llama-13b 2.606250
|
| 120 |
+
```
|
| 121 |
|
| 122 |
`big-refactor` branch:
|
| 123 |
|