Spaces:
Running
Running
Ludwig Stumpp
commited on
Commit
·
1d376a9
1
Parent(s):
2591e9a
Add links
Browse files- README.md +25 -24
- streamlit_app.py +25 -1
README.md
CHANGED
@@ -10,7 +10,7 @@ https://llm-leaderboard.streamlit.app/
|
|
10 |
|
11 |
We are always happy for contributions! You can contribute by the following:
|
12 |
|
13 |
-
- table work:
|
14 |
- filling missing entries
|
15 |
- adding a new model as a new row to the leaderboard and add the source of the evaluation to the sources table. Please keep alphabetic order.
|
16 |
- adding a new benchmark as a new column in the leaderboard and add the benchmark to the benchmarks table. Please keep alphabetic order.
|
@@ -20,35 +20,36 @@ We are always happy for contributions! You can contribute by the following:
|
|
20 |
|
21 |
## Leaderboard
|
22 |
|
23 |
-
| Model Name
|
24 |
-
|
|
25 |
-
| alpaca-13b
|
26 |
-
| cerebras-7b
|
27 |
-
| cerebras-13b
|
28 |
-
| chatglm-6b
|
29 |
-
| dolly-v2-12b
|
30 |
-
|
|
31 |
-
|
|
32 |
-
|
|
33 |
-
|
|
34 |
-
|
|
35 |
-
|
|
36 |
-
|
|
37 |
-
|
|
38 |
-
|
|
39 |
-
|
|
40 |
-
|
|
41 |
-
|
|
42 |
-
|
|
43 |
-
|
|
|
|
44 |
|
45 |
## Benchmarks
|
46 |
|
47 |
| Benchmark Name | Author | Link | Description |
|
48 |
| ----------------- | -------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
49 |
| Chatbot Arena Elo | LMSYS | https://lmsys.org/blog/2023-05-03-arena/ | "In this blog post, we introduce Chatbot Arena, an LLM benchmark platform featuring anonymous randomized battles in a crowdsourced manner. Chatbot Arena adopts the Elo rating system, which is a widely-used rating system in chess and other competitive games." (Source: https://lmsys.org/blog/2023-05-03-arena/) |
|
50 |
-
| LAMBADA | Paperno et al. | https://arxiv.org/abs/1606.06031
|
51 |
-
| TriviaQA | Joshi et al. | https://arxiv.org/abs/1705.03551v2
|
52 |
|
53 |
## Sources
|
54 |
|
|
|
10 |
|
11 |
We are always happy for contributions! You can contribute by the following:
|
12 |
|
13 |
+
- table work (don't forget the links):
|
14 |
- filling missing entries
|
15 |
- adding a new model as a new row to the leaderboard and add the source of the evaluation to the sources table. Please keep alphabetic order.
|
16 |
- adding a new benchmark as a new column in the leaderboard and add the benchmark to the benchmarks table. Please keep alphabetic order.
|
|
|
20 |
|
21 |
## Leaderboard
|
22 |
|
23 |
+
| Model Name | [Chatbot Arena Elo](https://lmsys.org/blog/2023-05-03-arena/) | [LAMBADA (zero-shot)](https://arxiv.org/abs/1606.06031) | [TriviaQA (zero-shot)](https://arxiv.org/abs/1705.03551v2 ) |
|
24 |
+
| -------------------------------------------------------------------------------------- | ------------------------------------------------------------- | ------------------------------------------------------- | ----------------------------------------------------------- |
|
25 |
+
| [alpaca-13b](https://crfm.stanford.edu/2023/03/13/alpaca.html) | [1008](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
26 |
+
| [cerebras-gpt-7b](https://huggingface.co/cerebras/Cerebras-GPT-6.7B) | | [0.636](https://www.mosaicml.com/blog/mpt-7b) | [0.141](https://www.mosaicml.com/blog/mpt-7b) |
|
27 |
+
| [cerebras-gpt-13b](https://huggingface.co/cerebras/Cerebras-GPT-13B) | | [0.635](https://www.mosaicml.com/blog/mpt-7b) | [0.146](https://www.mosaicml.com/blog/mpt-7b) |
|
28 |
+
| [chatglm-6b](https://chatglm.cn/blog) | [985](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
29 |
+
| [dolly-v2-12b](https://huggingface.co/databricks/dolly-v2-12b) | [944](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
30 |
+
| [eleuther-pythia-7b](https://huggingface.co/EleutherAI/pythia-6.9b) | | [0.667](https://www.mosaicml.com/blog/mpt-7b) | [0.198](https://www.mosaicml.com/blog/mpt-7b) |
|
31 |
+
| [eleuther-pythia-12b](https://huggingface.co/EleutherAI/pythia-12b) | | [0.704](https://www.mosaicml.com/blog/mpt-7b) | [0.233](https://www.mosaicml.com/blog/mpt-7b) |
|
32 |
+
| [fastchat-t5-3b](https://huggingface.co/lmsys/fastchat-t5-3b-v1.0) | [951](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
33 |
+
| [gpt-neox-20b](https://huggingface.co/EleutherAI/gpt-neox-20b) | | [0.719](https://www.mosaicml.com/blog/mpt-7b) | [0.347](https://www.mosaicml.com/blog/mpt-7b) |
|
34 |
+
| [gptj-6b](https://huggingface.co/EleutherAI/gpt-j-6b) | | [0.683](https://www.mosaicml.com/blog/mpt-7b) | [0.234](https://www.mosaicml.com/blog/mpt-7b) |
|
35 |
+
| [koala-13b](https://bair.berkeley.edu/blog/2023/04/03/koala/) | [1082](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
36 |
+
| [llama-7b](https://arxiv.org/abs/2302.13971) | | [0.738](https://www.mosaicml.com/blog/mpt-7b) | [0.443](https://www.mosaicml.com/blog/mpt-7b) |
|
37 |
+
| [llama-13b](https://arxiv.org/abs/2302.13971) | [932](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
38 |
+
| [mpt-7b](https://huggingface.co/mosaicml/mpt-7b) | | [0.702](https://www.mosaicml.com/blog/mpt-7b) | [0.343](https://www.mosaicml.com/blog/mpt-7b) |
|
39 |
+
| [oasst-pythia-12b](https://huggingface.co/OpenAssistant/pythia-12b-pre-v8-12.5k-steps) | [1065](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
40 |
+
| [opt-7b](https://huggingface.co/facebook/opt-6.7b) | | [0.677](https://www.mosaicml.com/blog/mpt-7b) | [0.227](https://www.mosaicml.com/blog/mpt-7b) |
|
41 |
+
| [opt-13b](https://huggingface.co/facebook/opt-13b) | | [0.692](https://www.mosaicml.com/blog/mpt-7b) | [0.282](https://www.mosaicml.com/blog/mpt-7b) |
|
42 |
+
| [stablelm-base-alpha-7b](https://huggingface.co/stabilityai/stablelm-base-alpha-7b) | | [0.533](https://www.mosaicml.com/blog/mpt-7b) | [0.049](https://www.mosaicml.com/blog/mpt-7b) |
|
43 |
+
| [stablelm-tuned-alpha-7b](https://huggingface.co/stabilityai/stablelm-tuned-alpha-7b) | [858](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
44 |
+
| [vicuna-13b](https://huggingface.co/lmsys/vicuna-13b-delta-v0) | [1169](https://lmsys.org/blog/2023-05-03-arena/) | | |
|
45 |
|
46 |
## Benchmarks
|
47 |
|
48 |
| Benchmark Name | Author | Link | Description |
|
49 |
| ----------------- | -------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
50 |
| Chatbot Arena Elo | LMSYS | https://lmsys.org/blog/2023-05-03-arena/ | "In this blog post, we introduce Chatbot Arena, an LLM benchmark platform featuring anonymous randomized battles in a crowdsourced manner. Chatbot Arena adopts the Elo rating system, which is a widely-used rating system in chess and other competitive games." (Source: https://lmsys.org/blog/2023-05-03-arena/) |
|
51 |
+
| LAMBADA | Paperno et al. | https://arxiv.org/abs/1606.06031 | "The LAMBADA evaluates the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse." (Source: https://huggingface.co/datasets/lambada) |
|
52 |
+
| TriviaQA | Joshi et al. | https://arxiv.org/abs/1705.03551v2 | "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions." (Source: https://arxiv.org/abs/1705.03551v2) |
|
53 |
|
54 |
## Sources
|
55 |
|
streamlit_app.py
CHANGED
@@ -1,6 +1,8 @@
|
|
|
|
|
|
|
|
1 |
import pandas as pd
|
2 |
import streamlit as st
|
3 |
-
import io
|
4 |
|
5 |
|
6 |
def extract_table_and_format_from_markdown_text(markdown_table: str) -> pd.DataFrame:
|
@@ -59,6 +61,27 @@ def extract_markdown_table_from_multiline(multiline: str, table_headline: str, n
|
|
59 |
return "".join(table)
|
60 |
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
def filter_dataframe(df: pd.DataFrame) -> pd.DataFrame:
|
63 |
"""
|
64 |
Adds a UI on top of a dataframe to let viewers filter columns
|
@@ -112,6 +135,7 @@ def setup_basic():
|
|
112 |
|
113 |
def setup_leaderboard(readme: str):
|
114 |
leaderboard_table = extract_markdown_table_from_multiline(readme, table_headline="## Leaderboard")
|
|
|
115 |
df_leaderboard = extract_table_and_format_from_markdown_text(leaderboard_table)
|
116 |
|
117 |
st.markdown("## Leaderboard")
|
|
|
1 |
+
import io
|
2 |
+
import re
|
3 |
+
|
4 |
import pandas as pd
|
5 |
import streamlit as st
|
|
|
6 |
|
7 |
|
8 |
def extract_table_and_format_from_markdown_text(markdown_table: str) -> pd.DataFrame:
|
|
|
61 |
return "".join(table)
|
62 |
|
63 |
|
64 |
+
def remove_markdown_links(text: str) -> str:
|
65 |
+
"""Modifies a markdown text to remove all markdown links.
|
66 |
+
Example: [DISPLAY](LINK) to DISPLAY
|
67 |
+
First find all markdown links with regex.
|
68 |
+
Then replace them with: $1
|
69 |
+
Args:
|
70 |
+
text (str): Markdown text containing markdown links
|
71 |
+
Returns:
|
72 |
+
str: Markdown text without markdown links.
|
73 |
+
"""
|
74 |
+
|
75 |
+
# find all markdown links
|
76 |
+
markdown_links = re.findall(r"\[([^\]]+)\]\(([^)]+)\)", text)
|
77 |
+
|
78 |
+
# remove link keep display text
|
79 |
+
for display, link in markdown_links:
|
80 |
+
text = text.replace(f"[{display}]({link})", display)
|
81 |
+
|
82 |
+
return text
|
83 |
+
|
84 |
+
|
85 |
def filter_dataframe(df: pd.DataFrame) -> pd.DataFrame:
|
86 |
"""
|
87 |
Adds a UI on top of a dataframe to let viewers filter columns
|
|
|
135 |
|
136 |
def setup_leaderboard(readme: str):
|
137 |
leaderboard_table = extract_markdown_table_from_multiline(readme, table_headline="## Leaderboard")
|
138 |
+
leaderboard_table = remove_markdown_links(leaderboard_table)
|
139 |
df_leaderboard = extract_table_and_format_from_markdown_text(leaderboard_table)
|
140 |
|
141 |
st.markdown("## Leaderboard")
|