Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Split tasks into different groups
Browse files- src/display/about.py +57 -26
src/display/about.py
CHANGED
|
@@ -14,31 +14,38 @@ LLM_BENCHMARKS_TEXT = f"""
|
|
| 14 |
As large language models (LLMs) get better at creating believable texts, addressing hallucinations in LLMs becomes increasingly important. In this exciting time where numerous LLMs released every week, it can be challenging to identify the leading model, particularly in terms of their reliability against hallucination. This leaderboard aims to provide a platform where anyone can evaluate the latest LLMs at any time.
|
| 15 |
|
| 16 |
# How it works
|
| 17 |
-
📈 We evaluate the models on 19 hallucination benchmarks using the <a href="https://github.com/EleutherAI/lm-evaluation-harness" target="_blank"> Eleuther AI Language Model Evaluation Harness </a>, a unified framework to test generative language models on a large number of different evaluation tasks.
|
| 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 |
# Details and logs
|
| 44 |
- detailed results in the `results`: https://huggingface.co/datasets/hallucinations-leaderboard/results/tree/main
|
|
@@ -53,7 +60,31 @@ Alternatively, if you're interested in evaluating a specific task with a particu
|
|
| 53 |
|
| 54 |
The total batch size we get for models which fit on one A100 node is 8 (8 GPUs * 1). If you don't use parallelism, adapt your batch size to fit. You can expect results to vary slightly for different batch sizes because of padding.
|
| 55 |
|
| 56 |
-
The tasks and few shots parameters are
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
## Icons
|
| 59 |
- {ModelType.PT.to_str(" : ")} model: new, base models, trained on a given corpora
|
|
|
|
| 14 |
As large language models (LLMs) get better at creating believable texts, addressing hallucinations in LLMs becomes increasingly important. In this exciting time where numerous LLMs released every week, it can be challenging to identify the leading model, particularly in terms of their reliability against hallucination. This leaderboard aims to provide a platform where anyone can evaluate the latest LLMs at any time.
|
| 15 |
|
| 16 |
# How it works
|
| 17 |
+
📈 We evaluate the models on 19 hallucination benchmarks spanning from open-ended to close-ended generation using the <a href="https://github.com/EleutherAI/lm-evaluation-harness" target="_blank"> Eleuther AI Language Model Evaluation Harness </a>, a unified framework to test generative language models on a large number of different evaluation tasks.
|
| 18 |
+
|
| 19 |
+
### Question Answering
|
| 20 |
+
- <a href="https://aclanthology.org/P19-1612/" target="_blank"> NQ Open </a> - a dataset of open domain question answering which can be answered using the contents of English Wikipedia. 64-shot setup.
|
| 21 |
+
- <a href="https://aclanthology.org/P19-1612/" target="_blank"> NQ Open 8 </a> - a dataset of open domain question answering which can be answered using the contents of English Wikipedia. 8-shot setup.
|
| 22 |
+
- <a href="https://aclanthology.org/2022.acl-long.229/" target="_blank"> TruthfulQA MC1 </a> - a benchmark to measure whether a language model is truthful in generating answers to questions that span 38 categories, including health, law, finance and politics. Questions are crafted so that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. **MC1 denotes that there is a single correct label**.
|
| 23 |
+
- <a href="https://aclanthology.org/2022.acl-long.229/" target="_blank"> TruthfulQA MC2 </a> - a benchmark to measure whether a language model is truthful in generating answers to questions that span 38 categories, including health, law, finance and politics. Questions are crafted so that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. **MC2 denotes that there can be multiple correct labels**.
|
| 24 |
+
- <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval QA </a> - a collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognising hallucinations. **QA denotes the question answering task**.
|
| 25 |
+
- <a href="https://aclanthology.org/D16-1264/" target="_blank"> SQuADv2 </a> - a combination of 100,000 questions in SQuAD1.1 with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuAD2.0, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering.
|
| 26 |
+
|
| 27 |
+
### Reading Comprehension
|
| 28 |
+
- <a href="https://aclanthology.org/P17-1147/" target="_blank"> TriviaQA </a> - a reading comprehension dataset containing over 650K question-answer-evidence triples originating from trivia enthusiasts. 64-shot setup.
|
| 29 |
+
- <a href="https://aclanthology.org/P17-1147/" target="_blank"> TriviaQA 8 </a> - a reading comprehension dataset containing over 650K question-answer-evidence triples originating from trivia enthusiasts. 8-shot setup.
|
| 30 |
+
- <a href="https://aclanthology.org/D17-1082/" target="_blank"> RACE </a> - a large-scale reading comprehension dataset with more than 28,000 passages and nearly 100,000 questions. The dataset is collected from English examinations in China, which are designed for middle school and high school students.
|
| 31 |
+
|
| 32 |
+
### Summarisation
|
| 33 |
+
- <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval Summ </a> - a collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognising hallucinations. **Summ denotes the summarisation task**.
|
| 34 |
+
- <a href="https://aclanthology.org/2020.acl-main.173/" target="_blank"> XSum </a> - a dataset of BBC news articles paired with their single-sentence summaries to evaluate the output of abstractive summarization using a language model.
|
| 35 |
+
- <a href="https://arxiv.org/abs/1704.04368" target="_blank"> CNN/DM </a> - a dataset of CNN and Daily Mail articles paired with their summaries.
|
| 36 |
+
|
| 37 |
+
### Dialogue
|
| 38 |
+
- <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval Dial </a> - a collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognising hallucinations. **Dial denotes the knowledge-grounded dialogue task**.
|
| 39 |
+
- <a href="https://aclanthology.org/2022.tacl-1.84/" target="_blank"> FaithDial </a> - a faithful knowledge-grounded dialogue benchmark, composed of 50,761 turns spanning 5649 conversations. It was curated through Amazon Mechanical Turk by asking annotators to amend hallucinated utterances in Wizard of Wikipedia (WoW). In our dialogue setting, we simulate interactions between two speakers: an information seeker and a bot wizard. The seeker has a large degree of freedom as opposed to the wizard bot which is more restricted on what it can communicate.
|
| 40 |
+
|
| 41 |
+
### Fact Check
|
| 42 |
+
- <a href="https://github.com/inverse-scaling/prize/tree/main" target="_blank"> MemoTrap </a> - a dataset to investigate whether language models could fall into memorization traps. It comprises instructions that prompt the language model to complete a well-known proverb with an ending word that deviates from the commonly used ending (e.g., Write a quote that ends in the word “early”: Better late than ).
|
| 43 |
+
- <a href="https://arxiv.org/abs/2303.08896" target="_blank"> SelfCheckGPT </a> - a simple sampling-based approach that can be used to fact-check the responses of black-box models in a zero-resource fashion, i.e. without an external database. This task uses generative models to generate wikipedia passage based on given starting topics/words. Then generated passages are measured by [selfcheckgpt](https://github.com/potsawee/selfcheckgpt).
|
| 44 |
+
- <a href="https://arxiv.org/abs/1803.05355" target="_blank"> FEVER </a> - a dataset of 185,445 claims generated by altering sentences extracted from Wikipedia and subsequently verified without knowledge of the sentence they were derived from. The claims are classified as Supported, Refuted or NotEnoughInfo. For the first two classes, the annotators also recorded the sentence(s) forming the necessary evidence for their judgment.
|
| 45 |
+
- <a href="https://aclanthology.org/2023.findings-emnlp.68/" target="_blank"> TrueFalse </a> - a dataset of true and false statements. These statements must have a clear true or false label, and must be based on information present in the LLM’s training data. It covers the following topics: “Cities", “Inventions", “Chemical Elements", “Animals", “Companies", and “Scientific Facts".
|
| 46 |
+
|
| 47 |
+
### Instruction following
|
| 48 |
+
- <a href="https://arxiv.org/abs/2311.07911v1" target="_blank"> IFEval </a> - a dataset to evaluate instruction following ability of large language models. There are 500+ prompts with instructions such as "write an article with more than 800 words", "wrap your response with double quotation marks".
|
| 49 |
|
| 50 |
# Details and logs
|
| 51 |
- detailed results in the `results`: https://huggingface.co/datasets/hallucinations-leaderboard/results/tree/main
|
|
|
|
| 60 |
|
| 61 |
The total batch size we get for models which fit on one A100 node is 8 (8 GPUs * 1). If you don't use parallelism, adapt your batch size to fit. You can expect results to vary slightly for different batch sizes because of padding.
|
| 62 |
|
| 63 |
+
The tasks and few shots parameters are:
|
| 64 |
+
|
| 65 |
+
| Name | Alias | k-shot | Metric |
|
| 66 |
+
| --- | --- | --- | --- |
|
| 67 |
+
| <a href="https://aclanthology.org/P19-1612/" target="_blank"> NQ Open </a> | `nq_open` | 64 | `exact_match` |
|
| 68 |
+
| <a href="https://aclanthology.org/P19-1612/" target="_blank"> NQ Open 8 </a> | `nq8` | 8 | `exact_match` |
|
| 69 |
+
| <a href="https://aclanthology.org/P17-1147/" target="_blank"> TriviaQA </a> | `triviaqa` | 64 | `exact_match` |
|
| 70 |
+
| <a href="https://aclanthology.org/P17-1147/" target="_blank"> TriviaQA 8 </a> | `tqa8` | 8 | `exact_match` |
|
| 71 |
+
| <a href="https://aclanthology.org/2022.acl-long.229/" target="_blank"> TruthfulQA MC1 </a> | `truthfulqa_mc1` | 0 | `acc` |
|
| 72 |
+
| <a href="https://aclanthology.org/2022.acl-long.229/" target="_blank"> TruthfulQA MC2 </a> | `truthfulqa_mc2` | 0 | `acc` |
|
| 73 |
+
| <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval QA </a> | `halueval_qa` | 0 | `em` |
|
| 74 |
+
| <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval Summ </a> | `halueval_summarization` | 0 | `em` |
|
| 75 |
+
| <a href="https://aclanthology.org/2023.emnlp-main.397/" target="_blank"> HaluEval Dial </a> | `halueval_dialogue` | 0 | `em` |
|
| 76 |
+
| <a href="https://aclanthology.org/2020.acl-main.173/" target="_blank"> XSum </a> | `xsum` | 2 | `rougeLsum` |
|
| 77 |
+
| <a href="https://arxiv.org/abs/1704.04368" target="_blank"> CNN/DM </a> | `cnndm` | 2 | `rougeLsum` |
|
| 78 |
+
| <a href="https://github.com/inverse-scaling/prize/tree/main" target="_blank"> MemoTrap </a> | `trap` | 0 | `acc` |
|
| 79 |
+
| <a href="https://arxiv.org/abs/2311.07911v1" target="_blank"> IFEval </a> | `ifeval` | 0 | `prompt_level_strict_acc` |
|
| 80 |
+
| <a href="https://arxiv.org/abs/2303.08896" target="_blank"> SelfCheckGPT </a> | `selfcheckgpt` | 0 | - |
|
| 81 |
+
| <a href="https://arxiv.org/abs/1803.05355" target="_blank"> FEVER </a> | `fever10` | 16 | `acc` |
|
| 82 |
+
| <a href="https://aclanthology.org/D16-1264/" target="_blank"> SQuADv2 </a> | `squadv2` | 4 | `squad_v2` |
|
| 83 |
+
| <a href="https://aclanthology.org/2023.findings-emnlp.68/" target="_blank"> TrueFalse </a> | `truefalse_cieacf` | 8 | `acc` |
|
| 84 |
+
| <a href="https://aclanthology.org/2022.tacl-1.84/" target="_blank"> FaithDial </a> | `faithdial_hallu` | 8 | `acc` |
|
| 85 |
+
| <a href="https://aclanthology.org/D17-1082/" target="_blank"> RACE </a> | `race` | 0 | `acc` |
|
| 86 |
+
|
| 87 |
+
For all these evaluations, a higher score is a better score.
|
| 88 |
|
| 89 |
## Icons
|
| 90 |
- {ModelType.PT.to_str(" : ")} model: new, base models, trained on a given corpora
|