weichiang commited on
Commit
7060483
β€’
1 Parent(s): 75e958e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +2 -2
app.py CHANGED
@@ -21,11 +21,11 @@ def make_leaderboard_md(elo_results):
21
  | [Vote](https://chat.lmsys.org/?arena) | [Blog](https://lmsys.org/blog/2023-05-03-arena/) | [GitHub](https://github.com/lm-sys/FastChat) | [Paper](https://arxiv.org/abs/2306.05685) | [Dataset](https://github.com/lm-sys/FastChat/blob/main/docs/dataset_release.md) | [Twitter](https://twitter.com/lmsysorg) | [Discord](https://discord.gg/HSWAKCrnFx) |
22
 
23
  πŸ† This leaderboard is based on the following three benchmarks.
24
- - [Chatbot Arena](https://chat.lmsys.org/?arena) - a crowdsourced, randomized battle platform. We use 90K+ user votes to compute Elo ratings.
25
  - [MT-Bench](https://arxiv.org/abs/2306.05685) - a set of challenging multi-turn questions. We use GPT-4 to grade the model responses.
26
  - [MMLU](https://arxiv.org/abs/2009.03300) (5-shot) - a test to measure a model's multitask accuracy on 57 tasks.
27
 
28
- πŸ’» Code: The Arena Elo ratings are computed by this [notebook]({notebook_url}). The MT-bench scores (single-answer grading on a scale of 10) are computed by [fastchat.llm_judge](https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge). The MMLU scores are mostly computed by [InstructEval](https://github.com/declare-lab/instruct-eval). Higher values are better for all benchmarks. Empty cells mean not available. Last updated: October, 2023.
29
  """
30
  return leaderboard_md
31
 
 
21
  | [Vote](https://chat.lmsys.org/?arena) | [Blog](https://lmsys.org/blog/2023-05-03-arena/) | [GitHub](https://github.com/lm-sys/FastChat) | [Paper](https://arxiv.org/abs/2306.05685) | [Dataset](https://github.com/lm-sys/FastChat/blob/main/docs/dataset_release.md) | [Twitter](https://twitter.com/lmsysorg) | [Discord](https://discord.gg/HSWAKCrnFx) |
22
 
23
  πŸ† This leaderboard is based on the following three benchmarks.
24
+ - [Chatbot Arena](https://chat.lmsys.org/?arena) - a crowdsourced, randomized battle platform. We use 100K+ user votes to compute Elo ratings.
25
  - [MT-Bench](https://arxiv.org/abs/2306.05685) - a set of challenging multi-turn questions. We use GPT-4 to grade the model responses.
26
  - [MMLU](https://arxiv.org/abs/2009.03300) (5-shot) - a test to measure a model's multitask accuracy on 57 tasks.
27
 
28
+ πŸ’» Code: The Arena Elo ratings are computed by this [notebook]({notebook_url}). The MT-bench scores (single-answer grading on a scale of 10) are computed by [fastchat.llm_judge](https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge). The MMLU scores are mostly computed by [InstructEval](https://github.com/declare-lab/instruct-eval). Higher values are better for all benchmarks. Empty cells mean not available. Last updated: November, 2023.
29
  """
30
  return leaderboard_md
31