📄 Seed Bench¶

SEED-Bench-1 consists of 19K multiple-choice questions with accurate human annotations, covering 12 evaluation dimensions including both the spatial and temporal understanding.

Tags: category:dataset

cards.seed_bench

type: TaskCard
loader: 
  type: LoadHF
  path: lmms-lab/SEED-Bench
preprocess_steps: 
  - type: ToImage
    field: image
    to_field: context
    process_every_value: True
  - type: ToRGB
    field: context
    process_every_value: True
  - type: ListFieldValues
    fields: 
      - choice_a
      - choice_b
      - choice_c
      - choice_d
    to_field: choices
  - type: Set
    fields: 
      context_type: video
  - type: MapValues
    mapping: 
      A: 0
      B: 1
      C: 2
      D: 3
    field: answer
task: tasks.qa.multiple_choice.with_context
templates: templates.qa.multiple_choice.with_context.no_intro.all
default_template: 
  type: MultipleChoiceTemplate
  input_format: "{context}\n{question}\n{choices}\nAnswer with the option's letter from the given choices directly."
  choices_separator: "\n"
  target_field: answer
  enumerator: capitals
[source]

Explanation about TaskCard¶

TaskCard delineates the phases in transforming the source dataset into model input, and specifies the metrics for evaluation of model output.

Args:

loader:
specifies the source address and the loading operator that can access that source and transform it into a unitxt multistream.

preprocess_steps:
list of unitxt operators to process the data source into model input.

task:
specifies the fields (of the already (pre)processed instance) making the inputs, the fields making the outputs, and the metrics to be used for evaluating the model output.

templates:
format strings to be applied on the input fields (specified by the task) and the output fields. The template also carries the instructions and the list of postprocessing steps, to be applied to the model output.

default_template:
a default template for tasks with very specific task dataset specific template

Explanation about MultipleChoiceTemplate¶

Formats the input that specifies a multiple-choice question, with a list of possible answers to choose from, and identifies the correct answer.

Args:

target_prefix (str): Optional prefix that can be added before the target label in
generated prompts or outputs.

choices_field (str): The key under which the multiple choices are stored in the
input and reference dictionaries.

target_field (str): The key under which the correct choice is stored in the
reference dictionary (can be integer index or textual label).

choices_separator (str): A string used to join formatted choices (e.g. “, “). source_choice_format (str): A Python format string used for displaying each choice

in the input fields (e.g. “{choice_numeral}. {choice_text}”).

target_choice_format (str): A Python format string used for displaying each choice
in the target or final output (e.g. “{choice_numeral}”).

enumerator (str): Determines how choice numerals are enumerated. Possible values
include “capitals”, “lowercase”, “numbers”, or “roman”.

shuffle_choices (bool): If True, shuffle the choices. The shuffling seed can be
set with shuffle_choices_seed.

shuffle_choices_seed (int, optional): If provided, the choices are shuffled with
this fixed integer seed for reproducibility.

sort_choices_by_length (bool): If True, sorts choices by their length (ascending). sort_choices_alphabetically (bool): If True, sorts choices in alphabetical order. reverse_choices (bool): If True, reverses the order of the choices after any

sorting has been applied. Defaults to False to preserve backward compatibility.

Explanation about ListFieldValues¶

Concatenates values of multiple fields into a list, and assigns it to a new field.

Explanation about Set¶

Sets specified fields in each instance, in a given stream or all streams (default), with specified values. If fields exist, updates them, if do not exist – adds them.

Args:
fields (Dict[str, object]): The fields to add to each instance. Use ‘/’ to access inner fields

use_deepcopy (bool) : Deep copy the input value to avoid later modifications

Examples:
# Set a value of a list consisting of “positive” and “negative” do field “classes” to each and every instance of all streams Set(fields={"classes": ["positive","negatives"]})

# In each and every instance of all streams, field “span” is to become a dictionary containing a field “start”, in which the value 0 is to be set Set(fields={"span/start": 0}

# In all instances of stream “train” only, Set field “classes” to have the value of a list consisting of “positive” and “negative” Set(fields={"classes": ["positive","negatives"], apply_to_stream=["train"]})

# Set field “classes” to have the value of a given list, preventing modification of original list from changing the instance. Set(fields={"classes": alist}), use_deepcopy=True) if now alist is modified, still the instances remain intact.

Explanation about LoadHF¶

Loads datasets from the HuggingFace Hub.

It supports loading with or without streaming, and it can filter datasets upon loading.
Args:

path:
The path or identifier of the dataset on the HuggingFace Hub.

name:
An optional dataset name.

data_dir:
Optional directory to store downloaded data.

split:
Optional specification of which split to load.

data_files:
Optional specification of particular data files to load.

revision:
Optional. The revision of the dataset. Often the commit id. Use in case you want to set the dataset version.

streaming (bool):
indicating if streaming should be used.

filtering_lambda (str, optional):
A lambda function for filtering the data after loading.

num_proc (int, optional):
Specifies the number of processes to use for parallel dataset loading.

Example:
Loading glue’s mrpc dataset
load_hf = LoadHF(path='glue', name='mrpc')

References: templates.qa.multiple_choice.with_context.no_intro.all, tasks.qa.multiple_choice.with_context