Dataset

MBJP

A development set constructed for this study from Java tasks in MBXP. It contains 17 functionally correct generated programs, each independently rated by two annotators. The target is their mean readability rating on a 1–5 scale.

The loader converts escaped newlines and quotes in the stored JSON into source text. Readability targets come from the two human ratings, not the high/low generation prompts.

Download dataset (JSON)

Samples 17

Samples

Click a sample ID to inspect its code, method scores, and BTTF feature diagnosis.


MBJP/1	3	0	0.215	0.188	0.383	23	11.67	14	0.464	58
MBJP/10	4.5	0.005	0.577	0.774	0.474	8	15.67	18	0.554	24
MBJP/11	4	0	0.544	0.487	0.522	14	17	18	0.574	36
MBJP/12	2	0	0.607	0.46	0.708	23	8.33	13	0.435	45
MBJP/14	4.5	0.695	0.971	0.794	0.482	4	17.67	18	0.568	24
MBJP/17	2	0.216	0.967	0.79	0.772	9	10.67	12	0.356	27
MBJP/18	4.5	0	0.742	0.504	0.617	14	14	16	0.591	35
MBJP/19	5	0.553	0.85	0.722	0.769	7	18	19	0.778	24
MBJP/2	4.5	0.009	0.467	0.716	0.605	8	17.33	18	0.607	25
MBJP/21	2.5	0.023	0.812	0.79	0.692	9	10.67	14.67	0.342	28
MBJP/22	5	0.474	0.909	0.689	0.862	7	19.33	19	0.754	25
MBJP/3	5	0.18	0.942	0.772	0.755	8	16.67	18	0.714	28
MBJP/4	3	0	0.732	0.636	0.404	11	17	18	0.459	28
MBJP/6	2.5	0.324	0.921	0.618	0.668	3	18	19	0.641	22
MBJP/7	3.5	0.019	0.797	0.592	0.408	7	10.67	17.67	0.548	24
MBJP/8	1	0	0.483	0.536	0.772	24	2.33	6.67	0.241	45
MBJP/9	1.5	0	0.337	0.399	0.711	30	4.67	8	0.299	55