DVBench / Benchmarks

Add a benchmark

dvbench/Coding

Code generation evaluations.

4 measures ↑ humaneval_0shot

dvbench/Commonsense

Commonsense completion and disambiguation evaluations.

2 measures ↑ hellaswag_10shot

dvbench/Knowledge

Academic and scientific knowledge evaluations.

4 measures ↑ mmlu_5shot

dvbench/Mathematics

Grade-school and competition mathematics evaluations.

2 measures ↑ math_4shot

dvbench/Multilingual

Multilingual exam, understanding, mathematics, and translation evaluations.

4 measures ↑ multilingual_understanding

dvbench/Reasoning

Multi-step and scientific reasoning evaluations.

2 measures ↑ bbh_3shot

dvbench/Tape dataset validation 20260905_201353

Two local JAX training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/20260905_201353/next_token_mse

dvbench/Tape dataset validation 20260906_004839

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_004839/next_token_mse

dvbench/Tape dataset validation 20260906_004900

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_004900/next_token_mse

dvbench/Tape dataset validation 20260906_004936

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_004936/next_token_mse

dvbench/Tape dataset validation 20260906_005049

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_005049/next_token_mse

dvbench/Tape dataset validation 20260906_005141

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_005141/next_token_mse

dvbench/Tape dataset validation 20260906_005256

Two local jax training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/jax_20260906_005256/next_token_mse

dvbench/Tape dataset validation 20260906_171316

Two local torch training runs for attestation and dataset-filter checks.

1 measure ↓ tape_validation/torch_20260906_171316/next_token_mse

dvbench/Truthfulness

Truthful response evaluation.

1 measure ↑ truthfulqa_0shot