README.md

metadata

license: apache-2.0
dataset_info:
  features:
    - name: relevant_lemmas
      sequence: string
    - name: prev_tactics
      sequence: string
    - name: context
      struct:
        - name: bg_goals
          list:
            - name: goal
              dtype: string
            - name: hypotheses
              sequence: string
        - name: fg_goals
          list:
            - name: goal
              dtype: string
            - name: hypotheses
              sequence: string
        - name: given_up_goals
          list:
            - name: goal
              dtype: string
            - name: hypotheses
              sequence: string
        - name: shelved_goals
          list:
            - name: goal
              dtype: string
            - name: hypotheses
              sequence: string
    - name: tactic
      dtype: string
  splits:
    - name: test
      num_bytes: 4006839384
      num_examples: 363042
    - name: train
      num_bytes: 22292111267
      num_examples: 1836339
  download_size: 170002248
  dataset_size: 26298950651
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
      - split: train
        path: data/train-*

Proverbot Scrapes

Here we include a dump of proofs in coq-gym using the proverbot9001 tool.

Dataset Structure:

• relevant_lemmas: A list of lemmas that are relevant to the proof. Each lemma is accompanied by its formal statement and registration in the proof context.
• prev_tactics: The tactics used in the previous steps of the proof (The first tactic being the lemma definition itself). This can help in understanding the sequence of operations leading to the current state.
• context: The current proof context, including background goals (bg_goals), foreground goals (fg_goals), given up goals (given_up_goals), and shelved goals (shelved_goals).
  • fg_goals: The primary goal(s) being addressed at the current step, along with any hypotheses that have been introduced.
• tactic: The tactic being applied in the current proof step.

Example Entry:

{
  'relevant_lemmas': [
    'Axiomatization2SequentCalculus_cpSC :  ClassicalPropositionalSequentCalculus L GammaD.',
    'reg_Axiomatization2SequentCalculus_cpSC :  RegisterClass P2D_reg (fun cpSC: unit => @Axiomatization2SequentCalculus_cpSC) 5.',
    ...
  ],
  'prev_tactics': [
    '\n\nLemma contrapositiveNP: forall (x y: expr),\n  |-- (~~ y --> x) --> (~~ x --> y).\n'
  ],
  'context': {
    'bg_goals': [],
    'fg_goals': [
      {
        'goal': 'forall x y : expr, provable (impp (impp (negp y) x) (impp (negp x) y))',
        'hypotheses': [
          'cpAX : ClassicalPropositionalLogic L Gamma',
          ...
        ]
      }
    ],
    'given_up_goals': [],
    'shelved_goals': []
  },
  'tactic': '\nProof.\n',
}