There are 24 examples not solved by any model.
Solving some of these can be a good signal that your model is indeed better than leading models if these are good problems.
CRUXEval-output/112, CRUXEval-output/113, CRUXEval-output/129, CRUXEval-output/177, CRUXEval-output/218, CRUXEval-output/229, CRUXEval-output/245, CRUXEval-output/250, CRUXEval-output/254, CRUXEval-output/259, CRUXEval-output/272, CRUXEval-output/280, CRUXEval-output/301, CRUXEval-output/307, CRUXEval-output/33, CRUXEval-output/340, CRUXEval-output/375, CRUXEval-output/44, CRUXEval-output/469, CRUXEval-output/488, CRUXEval-output/581, CRUXEval-output/622, CRUXEval-output/640, CRUXEval-output/671
| example_link | model | min_elo |
|---|---|---|
| CRUXEval-output/220 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/591 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/458 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/484 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/125 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/599 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/35 | gpt-4-turbo-2024-04-09+cot | 1508.173 |
| CRUXEval-output/5 | gpt-4-0613+cot | 1391.554 |
| CRUXEval-output/445 | gpt-4o+cot | 1301.760 |
| CRUXEval-output/149 | gpt-4o+cot | 1301.760 |
| CRUXEval-output/175 | gpt-4o+cot | 1301.760 |
| CRUXEval-output/163 | gpt-4o+cot | 1301.760 |
| CRUXEval-output/438 | gpt-4-0613 | 1277.414 |
| CRUXEval-output/556 | gpt-4-turbo-2024-04-09 | 1262.303 |
| CRUXEval-output/550 | gpt-3.5-turbo-0613+cot | 1116.648 |
| CRUXEval-output/568 | gpt-3.5-turbo-0613+cot | 1116.648 |
| CRUXEval-output/749 | codellama-34b+cot | 888.284 |
| CRUXEval-output/499 | mixtral-8x7b | 858.129 |
| CRUXEval-output/571 | codellama-7b+cot | 651.289 |
| CRUXEval-output/209 | phi-1 | 619.495 |
These are 10 problems with the lowest correlation with the overall evaluation (i.e. better models tend to do worse on these. )
| example_link | acc | tau |
|---|---|---|
| CRUXEval-output/329 | 0.641 | -0.412 |
| CRUXEval-output/563 | 0.744 | -0.410 |
| CRUXEval-output/118 | 0.462 | -0.329 |
| CRUXEval-output/297 | 0.333 | -0.324 |
| CRUXEval-output/333 | 0.385 | -0.275 |
| CRUXEval-output/691 | 0.308 | -0.237 |
| CRUXEval-output/150 | 0.308 | -0.233 |
| CRUXEval-output/132 | 0.436 | -0.228 |
| CRUXEval-output/209 | 0.026 | -0.226 |
| CRUXEval-output/57 | 0.615 | -0.194 |
Histogram of problems by the accuracy on each problem.
Histogram of problems by the minimum Elo to solve each problem.