Score tasks

[rimashahbazyan]

• Added SCORE: Systematic COnsistency and Robustness Evaluation for Large Language Models
• Fixed a bug for generate until tasks to default the "until" parameter to each model's end of sequence.

[CLAassistant]

All committers have signed the CLA.

[baberabb]

Hi! looks great mostly! Could you add the following changes:

1. run the pre-commit.

pip install pre-commit
pre-commit install
pre-commit run --all-files

2. add an entry to lm_eval/tasks/README.md describing the benchmark in 1 sentence as is done for the other entries in that table.

[rimashahbazyan]

Hi! looks great mostly! Could you add the following changes:

1. run the pre-commit.

pip install pre-commit
pre-commit install
pre-commit run --all-files

2. add an entry to lm_eval/tasks/README.md describing the benchmark in 1 sentence as is done for the other entries in that table.

Thanks!
Done with both!

[baberabb]

@rimashahbazyan Thanks! test failing due to a missing function in utils_agieval

AttributeError: module 'utils_agieval' has no attribute 'robustness_doc_to_text'

[rimashahbazyan]

@baberabb
Sorry, fixed :)

[rimashahbazyan]

@baberabb I did some small bugfixes, and double-checked everything, this PR is final; I won't commit anything to the branch, when do you think it could be reviewed?

[baberabb]

@baberabb I did some small bugfixes, and double-checked everything, this PR is final; I won't commit anything to the branch, when do you think it could be reviewed?

Sorry for the delay, I'll try merging by next week if thats ok. Wanted to verify no other task is affected by removing the default until.