Spaces:
Running
Running
File size: 18,776 Bytes
d9de755 697be1a 908b597 617d84c 2591e9a 24a15c0 f1e50f4 1d376a9 f1e50f4 c0dd25e 2591e9a f1e50f4 bc01ae8 908b597 5323497 908b597 c0dd25e 908b597 5323497 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 |
# 🏆 llm-leaderboard
A joint community effort to create one central leaderboard for LLMs. Contributions and corrections welcome!
## Interactive Dashboard
https://llm-leaderboard.streamlit.app/
## How to Contribute
We are always happy for contributions! You can contribute by the following:
- table work (don't forget the links):
- filling missing entries
- adding a new model as a new row to the leaderboard. Please keep alphabetic order.
- adding a new benchmark as a new column in the leaderboard and add the benchmark to the benchmarks table. Please keep alphabetic order.
- code work:
- improving the existing code
- requesting and implementing new features
## Leaderboard
| Model Name | Commercial Use? | Chatbot Arena Elo | HumanEval-Python (pass@1) | LAMBADA (zero-shot) | MMLU (zero-shot) | TriviaQA (zero-shot) |
| -------------------------------------------------------------------------------------- | --------------- | ------------------------------------------------ | ------------------------------------------------------------------------------ | --------------------------------------------- | ---------------- | --------------------------------------------- |
| [alpaca-13b](https://crfm.stanford.edu/2023/03/13/alpaca.html) | no | [1008](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [cerebras-gpt-7b](https://huggingface.co/cerebras/Cerebras-GPT-6.7B) | yes | | | [0.636](https://www.mosaicml.com/blog/mpt-7b) | 0.259 | [0.141](https://www.mosaicml.com/blog/mpt-7b) |
| [cerebras-gpt-13b](https://huggingface.co/cerebras/Cerebras-GPT-13B) | yes | | | [0.635](https://www.mosaicml.com/blog/mpt-7b) | 0.258 | [0.146](https://www.mosaicml.com/blog/mpt-7b) |
| [chatglm-6b](https://chatglm.cn/blog) | yes | [985](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [chinchilla-70b](https://arxiv.org/abs/2203.15556v1) | no | | | [0.774](https://arxiv.org/abs/2203.15556v1) | | |
| [code-cushman-001](https://arxiv.org/abs/2107.03374) | no | | [33.5](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [code-davinci-002](https://arxiv.org/abs/2207.10397v2) | yes | | [65.8](https://arxiv.org/abs/2207.10397v2) | | | |
| [codegen-16B-mono](https://huggingface.co/Salesforce/codegen-16B-mono) | yes | | [29.3](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [codegen-16B-multi](https://huggingface.co/Salesforce/codegen-16B-multi) | yes | | [18.3](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [codegx-13b](http://keg.cs.tsinghua.edu.cn/codegeex/) | no | | [22.9](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [codex-12b](https://arxiv.org/abs/2107.03374v2) | no | | [28.81](https://arxiv.org/abs/2107.03374v2) | | | |
| [dolly-v2-12b](https://huggingface.co/databricks/dolly-v2-12b) | yes | [944](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [eleuther-pythia-7b](https://huggingface.co/EleutherAI/pythia-6.9b) | yes | | | [0.667](https://www.mosaicml.com/blog/mpt-7b) | 0.265 | [0.198](https://www.mosaicml.com/blog/mpt-7b) |
| [eleuther-pythia-12b](https://huggingface.co/EleutherAI/pythia-12b) | yes | | | [0.704](https://www.mosaicml.com/blog/mpt-7b) | 0.253 | [0.233](https://www.mosaicml.com/blog/mpt-7b) |
| [fastchat-t5-3b](https://huggingface.co/lmsys/fastchat-t5-3b-v1.0) | yes | [951](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [gpt-3.5-175b](https://arxiv.org/abs/2303.08774v3) | yes | | [48.1](https://arxiv.org/abs/2303.08774v3) | [0.762](https://arxiv.org/abs/2303.08774v3) | | |
| [gpt-4](https://arxiv.org/abs/2303.08774v3) | yes | | [67.0](https://arxiv.org/abs/2303.08774v3) | | | |
| [gpt-neox-20b](https://huggingface.co/EleutherAI/gpt-neox-20b) | yes | | | [0.719](https://www.mosaicml.com/blog/mpt-7b) | 0.269 | [0.347](https://www.mosaicml.com/blog/mpt-7b) |
| [gpt-j-6b](https://huggingface.co/EleutherAI/gpt-j-6b) | yes | | | [0.683](https://www.mosaicml.com/blog/mpt-7b) | 0.261 | [0.234](https://www.mosaicml.com/blog/mpt-7b) |
| [koala-13b](https://bair.berkeley.edu/blog/2023/04/03/koala/) | no | [1082](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [llama-7b](https://arxiv.org/abs/2302.13971) | no | | [10.5](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | [0.738](https://www.mosaicml.com/blog/mpt-7b) | 0.302 | [0.443](https://www.mosaicml.com/blog/mpt-7b) |
| [llama-13b](https://arxiv.org/abs/2302.13971) | no | [932](https://lmsys.org/blog/2023-05-03-arena/) | [15.8](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [llama-33b](https://arxiv.org/abs/2302.13971) | no | | [21.7](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [llama-65b](https://arxiv.org/abs/2302.13971) | no | | [23.7](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [mpt-7b](https://huggingface.co/mosaicml/mpt-7b) | yes | | | [0.702](https://www.mosaicml.com/blog/mpt-7b) | 0.296 | [0.343](https://www.mosaicml.com/blog/mpt-7b) |
| [oasst-pythia-12b](https://huggingface.co/OpenAssistant/pythia-12b-pre-v8-12.5k-steps) | yes | [1065](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [opt-7b](https://huggingface.co/facebook/opt-6.7b) | no | | | [0.677](https://www.mosaicml.com/blog/mpt-7b) | 0.251 | [0.227](https://www.mosaicml.com/blog/mpt-7b) |
| [opt-13b](https://huggingface.co/facebook/opt-13b) | no | | | [0.692](https://www.mosaicml.com/blog/mpt-7b) | 0.257 | [0.282](https://www.mosaicml.com/blog/mpt-7b) |
| [palm-540b](https://arxiv.org/abs/2204.02311v5) | no | | [26.2](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | [0.779](https://arxiv.org/abs/2204.02311v5) | | |
| [stablelm-base-alpha-7b](https://huggingface.co/stabilityai/stablelm-base-alpha-7b) | yes | | | [0.533](https://www.mosaicml.com/blog/mpt-7b) | 0.251 | [0.049](https://www.mosaicml.com/blog/mpt-7b) |
| [stablelm-tuned-alpha-7b](https://huggingface.co/stabilityai/stablelm-tuned-alpha-7b) | no | [858](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
| [starcoder-base-16B](https://huggingface.co/bigcode/starcoderbase) | yes | | [30.4](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [starcoder-16B](https://huggingface.co/bigcode/starcoder) | yes | | [33.6](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [starcoder-16B (prompted)](https://huggingface.co/bigcode/starcoder) | yes | | [40.8](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | |
| [vicuna-13b](https://huggingface.co/lmsys/vicuna-13b-delta-v0) | no | [1169](https://lmsys.org/blog/2023-05-03-arena/) | | | | |
## Benchmarks
| Benchmark Name | Author | Link | Description |
| ----------------- | ---------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| 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/) |
| HumanEval | Chen et al. | https://arxiv.org/abs/2107.03374v2 | "It used to measure functional correctness for synthesizing programs from docstrings. It consists of 164 original programming problems, assessing language comprehension, algorithms, and simple mathematics, with some comparable to simple software interview questions." (Source: https://paperswithcode.com/dataset/humaneval) |
| 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) |
| MMLU | Hendrycks et al. | https://github.com/hendrycks/test | "The benchmark covers 57 subjects across STEM, the humanities, the social sciences, and more. It ranges in difficulty from an elementary level to an advanced professional level, and it tests both world knowledge and problem solving ability. Subjects range from traditional areas, such as mathematics and history, to more specialized areas like law and ethics. The granularity and breadth of the subjects makes the benchmark ideal for identifying a model’s blind spots." (Source: "https://paperswithcode.com/dataset/mmlu") |
| 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) |
## Sources
The results of this leaderboard are collected from the individual papers and published results of the model authors. For each reported value, the source is added as a link.
## Disclaimer
Above information may be wrong. If you want to use a published model for commercial use, please contact a lawyer. |