README.md

---
license: cc-by-sa-4.0
configs:
- config_name: bbh_logical_deduction_three_objects
  data_files:
  - split: test
    path: bbh_logical_deduction_three_objects/test-*
- config_name: bbh_navigate
  data_files:
  - split: test
    path: bbh_navigate/test-*
- config_name: bbh_object_counting
  data_files:
  - split: test
    path: bbh_object_counting/test-*
- config_name: drop
  data_files:
  - split: test
    path: drop/test-*
- config_name: gsm8k
  data_files:
  - split: test
    path: gsm8k/test-*
- config_name: hotpotqa
  data_files:
  - split: test
    path: hotpotqa/test-*
- config_name: mmlu_math
  data_files:
  - split: test
    path: mmlu_math/test-*
- config_name: multiarith
  data_files:
  - split: test
    path: multiarith/test-*
- config_name: singleop
  data_files:
  - split: test
    path: singleop/test-*
- config_name: singleq
  data_files:
  - split: test
    path: singleq/test-*
- config_name: squad
  data_files:
  - split: test
    path: squad/test-*
- config_name: svamp
  data_files:
  - split: test
    path: svamp/test-*
- config_name: tab_fact
  data_files:
  - split: test
    path: tab_fact/test-*
- config_name: vqa
  data_files:
  - split: test
    path: vqa/test-*
- config_name: winograd_wsc
  data_files:
  - split: test
    path: winograd_wsc/test-*
dataset_info:
- config_name: bbh_logical_deduction_three_objects
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: target
    dtype: string
  splits:
  - name: test
    num_bytes: 305159
    num_examples: 200
  download_size: 60084
  dataset_size: 305159
- config_name: bbh_navigate
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: target
    dtype: string
  splits:
  - name: test
    num_bytes: 166521
    num_examples: 200
  download_size: 29525
  dataset_size: 166521
- config_name: bbh_object_counting
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: target
    dtype: string
  splits:
  - name: test
    num_bytes: 128265
    num_examples: 200
  download_size: 31211
  dataset_size: 128265
- config_name: drop
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: section_id
    dtype: string
  - name: query_id
    dtype: string
  - name: passage
    dtype: string
  - name: question
    dtype: string
  - name: answers_spans
    struct:
    - name: spans
      sequence: string
    - name: types
      sequence: string
  splits:
  - name: test
    num_bytes: 957113
    num_examples: 250
  download_size: 469801
  dataset_size: 957113
- config_name: gsm8k
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: question
    dtype: string
  - name: answer
    dtype: string
  splits:
  - name: test
    num_bytes: 411558
    num_examples: 300
  download_size: 200727
  dataset_size: 411558
- config_name: hotpotqa
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: id
    dtype: string
  - name: question
    dtype: string
  - name: answer
    dtype: string
  - name: type
    dtype: string
  - name: level
    dtype: string
  - name: supporting_facts
    struct:
    - name: sent_id
      sequence: int64
    - name: title
      sequence: string
  - name: context
    struct:
    - name: sentences
      sequence:
        sequence: string
    - name: title
      sequence: string
  splits:
  - name: test
    num_bytes: 2163497
    num_examples: 250
  download_size: 1287407
  dataset_size: 2163497
- config_name: mmlu_math
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: question
    dtype: string
  - name: subject
    dtype: string
  - name: choices
    sequence: string
  - name: answer
    dtype: int64
  splits:
  - name: test
    num_bytes: 287231
    num_examples: 270
  download_size: 113739
  dataset_size: 287231
- config_name: multiarith
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: output_program
    dtype: string
  - name: output_answer
    dtype: string
  - name: split
    dtype: string
  - name: dataset
    dtype: string
  splits:
  - name: test
    num_bytes: 157371
    num_examples: 174
  download_size: 54214
  dataset_size: 157371
- config_name: singleop
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: output_program
    dtype: string
  - name: output_answer
    dtype: string
  - name: split
    dtype: string
  - name: dataset
    dtype: string
  splits:
  - name: test
    num_bytes: 118922
    num_examples: 159
  download_size: 45006
  dataset_size: 118922
- config_name: singleq
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: input
    dtype: string
  - name: output_program
    dtype: string
  - name: output_answer
    dtype: string
  - name: split
    dtype: string
  - name: dataset
    dtype: string
  splits:
  - name: test
    num_bytes: 96097
    num_examples: 109
  download_size: 39915
  dataset_size: 96097
- config_name: squad
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: id
    dtype: string
  - name: title
    dtype: string
  - name: context
    dtype: string
  - name: question
    dtype: string
  - name: answers
    struct:
    - name: answer_start
      sequence: int64
    - name: text
      sequence: string
  splits:
  - name: test
    num_bytes: 860040
    num_examples: 250
  download_size: 464857
  dataset_size: 860040
- config_name: svamp
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: ID
    dtype: string
  - name: Body
    dtype: string
  - name: Question
    dtype: string
  - name: Equation
    dtype: string
  - name: Answer
    dtype: string
  - name: Type
    dtype: string
  - name: question_concat
    dtype: string
  splits:
  - name: test
    num_bytes: 322658
    num_examples: 300
  download_size: 116772
  dataset_size: 322658
- config_name: tab_fact
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: id
    dtype: int64
  - name: table_id
    dtype: string
  - name: table_text
    dtype: string
  - name: table_caption
    dtype: string
  - name: statement
    dtype: string
  - name: label
    dtype: int64
  splits:
  - name: test
    num_bytes: 1137041
    num_examples: 200
  download_size: 475116
  dataset_size: 1137041
- config_name: vqa
  features:
  - name: cleaning_status
    dtype: string
  - name: image_path
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: 'null'
  - name: platinum_parsing_stratagy
    dtype: string
  - name: question_type
    dtype: string
  - name: multiple_choice_answer
    dtype: string
  - name: answers
    list:
    - name: answer
      dtype: string
    - name: answer_confidence
      dtype: string
    - name: answer_id
      dtype: int64
  - name: image_id
    dtype: int64
  - name: answer_type
    dtype: string
  - name: question_id
    dtype: int64
  - name: question
    dtype: string
  splits:
  - name: test
    num_bytes: 122801
    num_examples: 242
  download_size: 26070
  dataset_size: 122801
- config_name: winograd_wsc
  features:
  - name: cleaning_status
    dtype: string
  - name: platinum_prompt
    dtype: string
  - name: platinum_prompt_no_cot
    dtype: string
  - name: platinum_target
    sequence: string
  - name: original_target
    sequence: string
  - name: platinum_parsing_strategy
    dtype: string
  - name: text
    dtype: string
  - name: pronoun
    dtype: string
  - name: pronoun_loc
    dtype: int64
  - name: quote
    dtype: string
  - name: quote_loc
    dtype: int64
  - name: options
    sequence: string
  - name: label
    dtype: int64
  - name: source
    dtype: string
  splits:
  - name: test
    num_bytes: 198631
    num_examples: 200
  download_size: 54961
  dataset_size: 198631
task_categories:
- question-answering
language:
- en
---

# Dataset Card for PlatinumBench

[**🏆 Leaderboard**](http://platinum-bench.csail.mit.edu/)  |  [**🖥️ Code**](https://github.com/MadryLab/platinum-benchmarks/)  |  [**📖 Paper**](https://arxiv.org/abs/2502.03461)  |  [**🔍 Error Viewer**](http://platinum-bench.csail.mit.edu/inspect)

## Dataset Description

- **Homepage:** http://platinum-bench.csail.mit.edu/
- **Repository:** https://github.com/MadryLab/platinum-benchmarks/
- **Paper:** https://arxiv.org/abs/2502.03461
- **Leaderboard:** http://platinum-bench.csail.mit.edu/
- **Point of Contact:** [Joshua Vendrow](mailto:jvendrow@mit.edu), [Edward Vendrow](mailto:evendrow@mit.edu)

### Dataset Summary

_**Platinum Benchmarks**_ are benchmarks that are are carefully curated to minimize label errors and ambiguity, allowing us to measure reliability of models.

This dataset contains fifteen platinum benchmarks created by manually revising questions from existing datasets (see the github repo for details on accessing our revised subset of VQA). To revise each benchmark, we ran a variety of frontier models on individual examples and manually re-annotated any example for which at least one model made an error. See the paper for further details on the revision process.

### Load the Dataset

To load the dataset using HuggingFace `datasets`, you first need to `pip install datasets`, then run the following code:

```python
from datasets import load_dataset

ds = load_dataset("madrylab/platinum-bench", name="gsm8k", split="test") # or another subset
ds = ds.filter(lambda x: x['cleaning_status'] != 'rejected') # filter out rejected questions
```

## Dataset structure

### Dataset Subsets & Cleaning Statistics

Below we list each of the platinum benchmarks with the number of examples in each benchmark that we kept via consensus, revised, verified, or rejected. See "Data Fields" for a description of what each cleaning status means.

|         | Included |  |  |  | Excluded |
| ----- | ----- | ----- | ----- | ----- | ----- | 
Dataset | **# Included** | Consensus | Revised | Verified | Rejected
SingleOp (Platinum) | **150** | 142 | 0 | 8 | 9
SingleEq (Platinum) | **100** | 87 | 0 | 13 | 9
MultiArith (Platinum) | **171** | 165 | 3 | 3 | 3
SVAMP (Platinum) | **268** | 222 | 3 | 43 | 32
GSM8K (Platinum) | **271** | 227 | 1 | 43 | 29
MMLU High‑School Math (Platinum) | **268** | 106 | 0 | 162 | 2
Logic. Ded. 3-Obj (Platinum) | **200** | 199 | 0 | 1 | 0
Object Counting (Platinum) | **190** | 58 | 0 | 132 | 10
Navigate (Platinum) | **200** | 134 | 0 | 66 | 0
TabFact (Platinum) | **173** | 58 | 3 | 112 | 27
HotPotQA (Platinum) | **183** | 48 | 89 | 46 | 67
SQUAD2.0 (Platinum) | **164** | 78 | 43 | 43 | 86
DROP (Platinum) | **209** | 30 | 177 | 2 | 41
Winograd WSC (Platinum) | **195** | 77 | 0 | 118 | 5
VQA (Platinum) | **242** | 0 | 242 | 0 | 358

### Data Instances

We accessed each of the fourteen original natural language benchmarks that we revised from their respective huggingface repositories, and each benchmark had its own per-instance data fields/columns. We have standardized these benchmarks by providing pre-constructed prompts for each dataset (under 'platinum_prompt'). Each prompt template automatically formats the relevant dataset columns into a consistent structure. You can use these standardized prompts directly, but we include the original dataset columns for those interested in their own prompting, or to seamlessly subtitute our revised benchmarks for the original versions.

For VQA, we source images and annotataions from their [official website](https://visualqa.org/download.html), and reference images by their image path in the original downloaded directory format (see our GitHub repository for additional details).

An example from the PlatinumBench GSM8K subset looks as follows:
```
{'cleaning_status': 'consensus',
 'platinum_prompt': 'Solve the following math word problem.\n\nA robe takes 2 bolts of blue fiber and half that much white fiber.  How many bolts in total does it take?\n\nThink step-by-step. Then, provide the final answer as a single integer in the format "Answer: XXX" with no extra formatting.',
 'platinum_prompt_no_cot': 'Solve the following math word problem.\n\nA robe takes 2 bolts of blue fiber and half that much white fiber.  How many bolts in total does it take?\n\nThen, provide the final answer as a single integer in the format "Answer: XXX" with no extra formatting.',
 'platinum_target': ['3'],
 'platinum_parsing_strategy': 'math',
 'original_target': ['3']
 'question': 'A robe takes 2 bolts of blue fiber and half that much white fiber.  How many bolts in total does it take?',
 'answer': 'It takes 2/2=<<2/2=1>>1 bolt of white fiber\nSo the total amount of fabric is 2+1=<<2+1=3>>3 bolts of fabric\n#### 3'}
```

### Data Fields
- **cleaning_status** (`str`): One of:
	1. *concensus*: all LLMs agreed with the label, so the example was not manually reviewed (`platinum_target` == `original_target` by default).
	2. *verified*: the original target was maually verified to be correct (`platinum_target` == `original_target`).
	3. *revised*: the label is updated from the original label  (`platinum_target` != `original_target`).
	4. *rejected*: the example is removed due to issues such as ambiguity.
- **platinum_prompt** (`str`): A chain-of-thought question prompt that can be directly asked to a language model. This is constructed from fields in the original dataset.
- **platinum_prompt_no_cot** (`str`): The same prompt, but without explicity chain-of-thought instructions. This is used for models like `o1` that don't need chain-of-thought prompting.
- **platinum_target** (`List[str]`): The list of all correct answers for the question. In most cases there is just one correct answer.
- **original_target** (`str`): The original target provided in the dataset. This is can be different from the platinum target if it is incorrect.
- **platinum_parsing_strategy** (`str`): The parser that should be used to parse the LLM answer. Refer to the provided code.
- **image_path** (`str`): Only included for VQA. The image path from which to source the relevant image, such as: `'val2014/COCO_val2014_000000304481.jpg`.
- We also incude all the original dataset columns after these ones.

> [!NOTE]
> This HuggingFace dataset includes rejected questions that are not used for evaluation. To use only questions that we include in our platinum benchmarks, make sure to filter these out:
> 
>`ds = ds.filter(lambda x: x['cleaning_status'] != 'rejected')`

### Prompt Example

Here is an example of the standardized prompt we provide for a question from MultiArith:

```
Solve the following math word problem.

At the schools book fair Sam bought 13 adventure books and 17 mystery books. If 15 of the books were used, how many new books did he buy?

Think step-by-step. Then, provide the final answer as a single number in the format "Answer: XXX" with no extra formatting.
```

The specific prompt template and parsing strategy depends on the model, although many of them are common between datasets.

## Dataset Creation

### Curation Rationale

Many current LLM benchmarks are riddled with label noise such as mislabeled or ambiguous questions. Due to this label noise, progress in these benchmarks often stalls before models actually achieve reliable performance on them. As a result, the comminuty often considers these benchmarks to be "saturated" and discards them too early, discouraging machine learning practictioners from ever striving to achieve proper reliability. As a first step towards addressing this gap in benchmarking practices, we revise samples from fifteen "saturated" benchmark to minimize label noise.

### Source Data and Attribution
Each of the fifteen benchmarks that we revise was sourced from the following huggingface repositories:


|                       |  Type   | URL | Subset | Split | License
| -----                 | ------ | ----- | ---- | ----| ----|
| SingleOp              | Math  | https://huggingface.co/datasets/allenai/lila          | singleop                        | test       | [CC&nbsp;BY&nbsp;4.0](https://github.com/allenai/Lila/blob/main/LICENSE.txt)
| SingleEq              | Math  | https://huggingface.co/datasets/allenai/lila          | singleeq                        | test       | [CC&nbsp;BY&nbsp;4.0](https://github.com/allenai/Lila/blob/main/LICENSE.txt)
| MultiArith            | Math  | https://huggingface.co/datasets/allenai/lila          | multiarith                      | test       | [CC&nbsp;BY&nbsp;4.0](https://github.com/allenai/Lila/blob/main/LICENSE.txt)
| SVAMP                 | Math  | https://huggingface.co/datasets/ChilleD/svamp         | default                         | test       | [MIT](https://github.com/arkilpatel/SVAMP/blob/main/LICENSE)
| GSM8K                 | Math  | https://huggingface.co/datasets/openai/gsm8k          | main                            | test       | [MIT](https://github.com/openai/grade-school-math/blob/master/LICENSE)
| MMLU&nbsp;High‑School&nbsp;Math | Math  | https://huggingface.co/datasets/cais/mmlu             | high_school_mathematics         | test       | [MIT](https://github.com/hendrycks/test/blob/master/LICENSE)
| Logic.&nbsp;Ded.&nbsp;3-Obj     | Logic | https://huggingface.co/datasets/maveriq/bigbenchhard  | logical_deduction_three_objects | train      | [MIT](https://github.com/suzgunmirac/BIG-Bench-Hard/blob/main/LICENSE)
| Object Counting       | Logic | https://huggingface.co/datasets/maveriq/bigbenchhard  | object_counting                 | train      | [MIT](https://github.com/suzgunmirac/BIG-Bench-Hard/blob/main/LICENSE)
| Navigate              | Logic | https://huggingface.co/datasets/maveriq/bigbenchhard  | navigate                        | train      | [MIT](https://github.com/suzgunmirac/BIG-Bench-Hard/blob/main/LICENSE)
| TabFact               | Table&nbsp;Understanding   | https://huggingface.co/datasets/wenhu/tab_fact        | tab_fact                        | test       | [CC&nbsp;BY&nbsp;4.0](https://creativecommons.org/licenses/by/4.0/legalcode)
| HotPotQA             | Reading&nbsp;Comp.    | https://huggingface.co/datasets/hotpotqa/hotpot_qa     | distractor                      | validation | [CC&nbsp;BY‑SA&nbsp;4.0](https://creativecommons.org/licenses/by-sa/4.0/legalcode)
| SQuAD2.0              | Reading&nbsp;Comp.    | https://huggingface.co/datasets/rajpurkar/squad_v2    | squad_v2                        | validation | [CC&nbsp;BY‑SA&nbsp;4.0](https://creativecommons.org/licenses/by-sa/4.0/legalcode)
| DROP                  | Reading&nbsp;Comp.    | https://huggingface.co/datasets/ucinlp/drop           | default                         | validation | [CC&nbsp;BY‑SA&nbsp;4.0](https://creativecommons.org/licenses/by-sa/4.0/legalcode)
| Wingograd WSC         | Commonsense    | https://huggingface.co/datasets/ErnestSDavis/winograd_wsc  | wsc285 | test | [CC&nbsp;BY&nbsp;4.0](https://creativecommons.org/licenses/by/4.0/legalcode)
| VQA                            | Vision     | https://visualqa.org/download.html                                         | N/A       |    validation | [CC&nbsp;BY&nbsp;4.0](https://creativecommons.org/licenses/by/4.0/legalcode)

Please defer to the datasets cards of these benchmarks for further details on their collection and annotation process.


## Additional Information

### Licensing Information

See the table above for the licensing information of the original datasets upon which our work is based. The further annotations we provide are licensed under the [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/legalcode) license.

### Citation Information
Cite this dataset and the source datasets (see [sources.bib](https://github.com/MadryLab/platinum-benchmarks/blob/main/sources.bib)).

```
@misc{vendrow2025largelanguagemodelbenchmarks,
      title={Do Large Language Model Benchmarks Test Reliability?}, 
      author={Joshua Vendrow and Edward Vendrow and Sara Beery and Aleksander Madry},
      year={2025},
      eprint={2502.03461},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2502.03461}, 
}
```