Social Foundations of Computation

benchbench

BenchBench is a Python package to evaluate multi-task benchmarks.

causal-features

Code to reproduce the paper "Do causal predictors generalize better to new domains?"

error-parity

Achieve error-rate fairness between societal groups for any score-based classifier.

folktables

Datasets derived from US census data

folktexts

Get classification risk scores on tabular tasks using LLMs

lawma

Lawma: A lightly fine-tuned Llama model for legal classification tasks.

surveying-language-models

Code to reproduce the paper "Questioning the Survey Responses of Large Language Models"

training-on-the-test-task

Code to reproduce the experiments in the paper Training on the Test Task Confounds Evaluation and Emergence.

tttlm

Test-time-training on nearest neighbors for large language models

whynot

A Python sandbox for decision making in dynamics

socialfoundations/whynot

A Python sandbox for decision making in dynamics

socialfoundations/folktables

Datasets derived from US census data

socialfoundations/tttlm

Test-time-training on nearest neighbors for large language models

socialfoundations/error-parity

Achieve error-rate fairness between societal groups for any score-based classifier.

socialfoundations/folktexts

Get classification risk scores on tabular tasks using LLMs

socialfoundations/lawma

Lawma: A lightly fine-tuned Llama model for legal classification tasks.

socialfoundations/benchbench

BenchBench is a Python package to evaluate multi-task benchmarks.

socialfoundations/surveying-language-models

Code to reproduce the paper "Questioning the Survey Responses of Large Language Models"

socialfoundations/training-on-the-test-task

Code to reproduce the experiments in the paper Training on the Test Task Confounds Evaluation and Emergence.

socialfoundations/causal-features

Code to reproduce the paper "Do causal predictors generalize better to new domains?"

socialfoundations/backward_baselines

Code for "Is your model predicting the past?"

socialfoundations/lm-evaluation-harness

A framework for few-shot evaluation of language models.

socialfoundations/twitter-predictability