Datasets:

cs-giung
/

base-evals

id stringlengths 25 25	context stringlengths 7 24	choices listlengths 4 4	label int64 0 3
tiny_arithmark_2_0:000001	59 + 45 =	[ " 100", " 104", " 98", " 92" ]	1
tiny_arithmark_2_0:000002	0 + 83 =	[ " 83", " 85", " 73", " 68" ]	0
tiny_arithmark_2_0:000006	(132 / 12) + (46 - 15) =	[ " 45", " 42", " 34", " 62" ]	1
tiny_arithmark_2_0:000008	2 * 12 =	[ " 28", " 18", " 44", " 24" ]	3
tiny_arithmark_2_0:000011	85 - 16 =	[ " 69", " 65", " 79", " 57" ]	0
tiny_arithmark_2_0:000013	1 * 11 + 26 =	[ " 39", " 42", " 49", " 37" ]	3
tiny_arithmark_2_0:000014	22 - 15 =	[ " 5", " 7", " 1", " 27" ]	1
tiny_arithmark_2_0:000026	16 + 88 =	[ " 100", " 104", " 98", " 84" ]	1
tiny_arithmark_2_0:000029	90 - 44 =	[ " 49", " 51", " 58", " 46" ]	3
tiny_arithmark_2_0:000032	64 - 51 =	[ " 13", " 15", " 5", " 33" ]	0
tiny_arithmark_2_0:000038	57 + 4 =	[ " 61", " 57", " 67", " 41" ]	0
tiny_arithmark_2_0:000039	9 * 4 + 15 / 5 =	[ " 42", " 39", " 33", " 27" ]	1
tiny_arithmark_2_0:000047	(6 / 3) + 11 =	[ " 14", " 13", " 21", " 33" ]	1
tiny_arithmark_2_0:000050	55 - 36 =	[ " 15", " 29", " 7", " 19" ]	3
tiny_arithmark_2_0:000060	112 - 105 =	[ " 6", " 13", " 7", " 27" ]	2
tiny_arithmark_2_0:000079	4 * (7 + 26 - 29) =	[ " 20", " 22", " 16", " 31" ]	2
tiny_arithmark_2_0:000086	3 * (60 - 46) =	[ " 43", " 47", " 42", " 27" ]	2
tiny_arithmark_2_0:000094	53 + 21 =	[ " 74", " 77", " 66", " 54" ]	0
tiny_arithmark_2_0:000096	82 + 48 =	[ " 130", " 134", " 138", " 110" ]	0
tiny_arithmark_2_0:000101	3 * 1 =	[ " 1", " 3", " 8", " 15" ]	1
tiny_arithmark_2_0:000102	91 - 36 + 1 * 1 =	[ " 57", " 51", " 36", " 56" ]	3
tiny_arithmark_2_0:000108	24 - 5 + 14 =	[ " 34", " 33", " 38", " 53" ]	1
tiny_arithmark_2_0:000113	(37 + 35) + (132 / 11) =	[ " 88", " 94", " 64", " 84" ]	3
tiny_arithmark_2_0:000122	33 - 3 * 6 =	[ " 14", " 10", " 15", " 0" ]	2
tiny_arithmark_2_0:000128	7 * (23 + 17 - 29) =	[ " 78", " 77", " 71", " 62" ]	1
tiny_arithmark_2_0:000130	(22 + 35) * 3 =	[ " 169", " 181", " 186", " 171" ]	3
tiny_arithmark_2_0:000131	20 / 4 + 5 - 9 =	[ " 3", " 9", " 16", " 1" ]	3
tiny_arithmark_2_0:000147	30 / 5 =	[ " 8", " 14", " 6", " 18" ]	2
tiny_arithmark_2_0:000149	(24 - 7 + 7) * 1 =	[ " 25", " 24", " 19", " 39" ]	1
tiny_arithmark_2_0:000156	90 / 10 + 54 - 21 =	[ " 41", " 32", " 42", " 62" ]	2
tiny_arithmark_2_0:000158	(53 + 6) - (7 + 26) =	[ " 26", " 23", " 20", " 6" ]	0
tiny_arithmark_2_0:000162	100 / 10 - 0 =	[ " 8", " 16", " 22", " 10" ]	3
tiny_arithmark_2_0:000163	8 + 1 * 9 - 16 =	[ " 4", " 11", " 13", " 1" ]	3
tiny_arithmark_2_0:000171	7 * 6 =	[ " 42", " 41", " 37", " 27" ]	0
tiny_arithmark_2_0:000177	(12 - 12) * 6 =	[ " 2", " 8", " 20", " 0" ]	3
tiny_arithmark_2_0:000180	19 + 7 * 7 =	[ " 70", " 74", " 80", " 68" ]	3
tiny_arithmark_2_0:000181	5 * 12 + 8 =	[ " 66", " 76", " 68", " 53" ]	2
tiny_arithmark_2_0:000187	2 * (51 - 2) =	[ " 95", " 108", " 98", " 86" ]	2
tiny_arithmark_2_0:000192	82 + 36 =	[ " 120", " 123", " 118", " 133" ]	2
tiny_arithmark_2_0:000205	44 + 4 * 10 - 50 =	[ " 32", " 26", " 34", " 54" ]	2
tiny_arithmark_2_0:000207	5 + 48 - 22 + 12 =	[ " 39", " 37", " 23", " 43" ]	3
tiny_arithmark_2_0:000208	8 * 3 + 48 / 8 =	[ " 28", " 36", " 15", " 30" ]	3
tiny_arithmark_2_0:000218	4 * (63 - 46) =	[ " 66", " 60", " 68", " 80" ]	2
tiny_arithmark_2_0:000221	55 + 44 - 32 =	[ " 65", " 77", " 67", " 79" ]	2
tiny_arithmark_2_0:000229	10 * 7 + 99 / 9 =	[ " 81", " 82", " 71", " 101" ]	0
tiny_arithmark_2_0:000232	92 - 23 =	[ " 66", " 75", " 69", " 57" ]	2
tiny_arithmark_2_0:000234	7 * (56 - 54) =	[ " 10", " 4", " 34", " 14" ]	3
tiny_arithmark_2_0:000237	(20 - 14) * 7 =	[ " 39", " 36", " 42", " 22" ]	2
tiny_arithmark_2_0:000239	46 - 5 =	[ " 43", " 41", " 31", " 61" ]	1
tiny_arithmark_2_0:000240	17 + 59 =	[ " 73", " 76", " 82", " 56" ]	1
tiny_arithmark_2_0:000241	10 / 2 =	[ " 4", " 5", " 15", " 25" ]	1
tiny_arithmark_2_0:000248	1 * (57 - 1) =	[ " 53", " 51", " 56", " 76" ]	2
tiny_arithmark_2_0:000249	2 * (3 + 30 - 10) =	[ " 44", " 36", " 26", " 46" ]	3
tiny_arithmark_2_0:000260	84 + 38 =	[ " 118", " 114", " 137", " 122" ]	3
tiny_arithmark_2_0:000262	(16 + 23) * 3 + 7 =	[ " 124", " 120", " 119", " 144" ]	0
tiny_arithmark_2_0:000271	97 - 27 =	[ " 69", " 78", " 70", " 50" ]	2
tiny_arithmark_2_0:000276	0 / 1 + 56 =	[ " 60", " 46", " 36", " 56" ]	3
tiny_arithmark_2_0:000278	74 - 9 =	[ " 61", " 60", " 53", " 65" ]	3
tiny_arithmark_2_0:000280	83 + 11 =	[ " 94", " 98", " 100", " 106" ]	0
tiny_arithmark_2_0:000281	41 + 65 =	[ " 107", " 114", " 106", " 94" ]	2
tiny_arithmark_2_0:000283	55 + 60 - 24 =	[ " 93", " 91", " 81", " 106" ]	1
tiny_arithmark_2_0:000284	65 - 3 * 7 =	[ " 41", " 36", " 44", " 56" ]	2
tiny_arithmark_2_0:000286	37 + 9 * 2 =	[ " 58", " 47", " 43", " 55" ]	3
tiny_arithmark_2_0:000290	55 + 15 - 2 =	[ " 71", " 60", " 83", " 68" ]	3
tiny_arithmark_2_0:000292	(70 - 51) * 6 =	[ " 116", " 114", " 109", " 126" ]	1
tiny_arithmark_2_0:000297	10 * 1 + 84 / 12 =	[ " 14", " 27", " 17", " 37" ]	2
tiny_arithmark_2_0:000300	47 + 75 =	[ " 124", " 132", " 102", " 122" ]	3
tiny_arithmark_2_0:000301	16 / 8 =	[ " 2", " 1", " 8", " 22" ]	0
tiny_arithmark_2_0:000302	0 + 34 =	[ " 35", " 42", " 34", " 49" ]	2
tiny_arithmark_2_0:000308	5 + 59 =	[ " 64", " 61", " 72", " 49" ]	0
tiny_arithmark_2_0:000322	12 / 1 =	[ " 13", " 12", " 22", " 27" ]	1
tiny_arithmark_2_0:000326	5 + 5 * 8 =	[ " 45", " 48", " 50", " 25" ]	0
tiny_arithmark_2_0:000328	22 + 2 * 4 =	[ " 30", " 31", " 20", " 50" ]	0
tiny_arithmark_2_0:000335	2 * (16 + 7 - 3) =	[ " 38", " 40", " 32", " 28" ]	1
tiny_arithmark_2_0:000347	0 * (29 + 14) =	[ " 4", " 5", " 15", " 0" ]	3
tiny_arithmark_2_0:000350	91 - 26 =	[ " 62", " 75", " 65", " 45" ]	2
tiny_arithmark_2_0:000356	10 * 2 + 21 =	[ " 41", " 44", " 31", " 56" ]	0
tiny_arithmark_2_0:000360	(3 + 38) + (8 / 2) =	[ " 44", " 45", " 39", " 65" ]	1
tiny_arithmark_2_0:000372	69 + 9 =	[ " 82", " 86", " 78", " 63" ]	2
tiny_arithmark_2_0:000378	58 + 27 =	[ " 84", " 95", " 85", " 65" ]	2
tiny_arithmark_2_0:000379	21 / 7 =	[ " 4", " 13", " 23", " 3" ]	3
tiny_arithmark_2_0:000382	45 + 94 =	[ " 139", " 140", " 134", " 124" ]	0
tiny_arithmark_2_0:000383	27 + 4 * 10 - 30 =	[ " 40", " 43", " 37", " 22" ]	2
tiny_arithmark_2_0:000386	(57 + 32) - (3 + 1) =	[ " 85", " 82", " 79", " 97" ]	0
tiny_arithmark_2_0:000396	(5 + 10) * (15 - 11) =	[ " 62", " 60", " 65", " 48" ]	1
tiny_arithmark_2_0:000397	74 + 79 =	[ " 157", " 163", " 153", " 141" ]	2
tiny_arithmark_2_0:000398	33 + 12 * 5 =	[ " 89", " 99", " 93", " 108" ]	2
tiny_arithmark_2_0:000401	3 * 6 + 30 / 5 =	[ " 28", " 24", " 32", " 12" ]	1
tiny_arithmark_2_0:000403	94 + 51 =	[ " 145", " 148", " 155", " 133" ]	0
tiny_arithmark_2_0:000407	30 / 5 - 3 =	[ " 3", " 2", " 11", " 18" ]	0
tiny_arithmark_2_0:000409	90 - 36 + 4 * 10 =	[ " 91", " 86", " 94", " 106" ]	2
tiny_arithmark_2_0:000413	8 * 9 + 8 - 52 =	[ " 25", " 33", " 16", " 28" ]	3
tiny_arithmark_2_0:000418	8 * (16 + 0) =	[ " 125", " 128", " 122", " 113" ]	1
tiny_arithmark_2_0:000419	(30 - 0) * 4 =	[ " 117", " 126", " 120", " 132" ]	2
tiny_arithmark_2_0:000421	66 - 31 + 40 =	[ " 71", " 69", " 75", " 95" ]	2
tiny_arithmark_2_0:000424	117 - 1 =	[ " 112", " 126", " 136", " 116" ]	3
tiny_arithmark_2_0:000435	59 - 32 + 20 =	[ " 44", " 53", " 32", " 47" ]	3
tiny_arithmark_2_0:000438	60 / 6 =	[ " 11", " 10", " 0", " 25" ]	1
tiny_arithmark_2_0:000442	113 - 99 =	[ " 10", " 14", " 8", " 34" ]	1
tiny_arithmark_2_0:000444	(39 + 57) - (5 + 26) =	[ " 68", " 60", " 45", " 65" ]	3

End of preview. Expand in Data Studio

base-evals

0-shot, 4-choice continuation evaluation for pretrained base models.

Splits

split	source	source split	rows
`tiny_arithmark_2_0`	`AxiomicLabs/ArithMark-2.0@5bd986c7`	`train`	500
`tiny_arithmark_3_0`	`AxiomicLabs/Arithmark-3.0@6f6e59dd`	`train`	500
`tiny_hellaswag`	`Rowan/hellaswag@218ec52e`	`validation`	500
`tiny_logicmark`	`AxiomicLabs/LogicMark@e357c145`	`train`	500
`tiny_openbookqa`	`allenai/openbookqa@388097ea`	`test`	500
`tiny_pythonio`	`XiangPan/pythonio-mc@a85256a1`	`test`	500
`tiny_truthfulqa`	`EleutherAI/truthful_qa_mc@b6468d07`	`validation`	500
total			3,500

Format

id: string, context: string, choices: list[string], label: int64. Labels are zero-based. Input is context + choices[i], with exactly one ASCII space before each choice and no trailing context whitespace. PythonIO uses LF line endings, Input: {input}\nOutput: contexts, and choices without outer trailing whitespace; code indentation and output-value whitespace are retained. HellaSwag uses context cleanup; TruthfulQA uses Q: {question}\nA:.

Evaluation

No few-shot examples, chat template, CoT, or added BOS/EOS.

acc: choose the largest sum of continuation-token log-probabilities.
acc_norm: choose the largest mean log-probability per continuation token.

Exclude context/padding; include boundary whitespace. Reject ambiguous token boundaries and overlength inputs. Break ties by first index. Report each split separately. This custom protocol is not HellaSwag character normalization, TruthfulQA MC2, or IRT.

Provenance

Each split selects 500 questions with random.Random(42); OpenBookQA retains its full test set. ArithMark 3.0 excludes time_session prompts without duration information before sampling. Rows with fewer than four choices are excluded; larger sets keep the answer and three sampled incorrect choices. Source row/choice order is preserved. Full rules, hashes and indices: manifest.json; original inputs/cards: sources/; offline rebuild: build.py.

Evaluation only, including upstream train splits. Original source terms apply; PythonIO and OpenBookQA have no declared license. Pretraining contamination is unknown.

Downloads last month: 180