Datasets:
Modalities:
Text
Formats:
parquet
Size:
< 1K
Tags:
degen_test_1_countdown
budget-32k
iterative
hosted_vllm-Qwen-Qwen3-4B-Thinking-2507<api_base>http:--10.32.36.199:9080-v1
smart-termination
max-rounds-50
License:
question stringlengths 129 133 | metadata stringlengths 136 144 | task_source stringclasses 1
value | formatted_prompt listlengths 2 2 | responses_by_sample listlengths 1 1 | tokens_by_sample listlengths 1 1 | num_rounds_by_sample listlengths 1 1 | continuation_prompts_by_sample listlengths 1 1 | continuation_text_by_sample listlengths 1 1 | strategy_type stringclasses 1
value | termination_reason_by_sample listlengths 1 1 | correct_round_by_sample listlengths 1 1 | smart_termination_enabled bool 1
class | answers_by_sample listlengths 1 1 | evals_by_sample listlengths 1 1 | eval_metadata_by_sample listlengths 1 1 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Using the numbers 99, 83, 4, 15, 38, 56, create an expression that equals 666. You may use +, -, *, / and each number at most once. | {"numbers": [99, 83, 4, 15, 38, 56], "target": 666, "operators_used": ["-", "/", "+", "*", "-"], "expression": "99 - 83 / 4 + 15 * 38 - 56"} | countdown | [
{
"content": "You are a helpful assistant that solves math problems by reasoning step-by-step. Show your work clearly and provide complete solutions. When you have a final answer, wrap it in <answer></answer> tags.",
"role": "system"
},
{
"content": "Using the numbers 99, 83, 4, 15, 38, 56, create a... | [
[
"<think>\nOkay, let's try to figure out how to make 666 using the numbers 99, 83, 4, 15, 38, 56. Each number can be used at most once, and we can use +, -, *, /. Hmm, 666 is a pretty big number, so multiplication might be involved here. Let me think.\n\nFirst, let's list out the numbers: 99, 83, 4, 15, 38, 56... | [
32831
] | [
30
] | [
[
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actually wrong, try again.",
"This is actuall... | [
[
"<think>\nOkay, let's try to figure out how to make 666 using the numbers 99, 83, 4, 15, 38, 56. Each number can be used at most once, and we can use +, -, *, /. Hmm, 666 is a pretty big number, so multiplication might be involved here. Let me think.\n\nFirst, let's list out the numbers: 99, 83, 4, 15, 38, 56... | iterative | [
null
] | [
null
] | true | [
[
[
"38*(15 + (99 - 83)/4) - 56"
],
[
"(38 * (15 + (99 - 83) / 4)) - 56"
],
[
"(38 * (15 + (99 - 83) / 4)) - 56",
"(38*(15 + (99 - 83)/4)) - 56",
"(38*(15 + (99 - 83)/4)) - 56"
],
[
"38*(15 + (99 - 83)/4) - 56"
],
[
"38*(15 + (99 - 83)/4)... | [
[
[
true
],
[
true
],
[
true,
true,
true
],
[
true
],
[
true
],
[
true
],
[
true,
true
],
[
true
],
[
true,
true
],
[
true,
true
],
[
tr... | [
[
[
{
"answer_block": "38*(15 + (99 - 83)/4) - 56",
"error": null,
"final_answer": 666,
"is_correct": true
}
],
[
{
"answer_block": "(38 * (15 + (99 - 83) / 4)) - 56",
"error": null,
"final_answer": 666,
"is_correct": true... |
"Using the numbers 74, 74, 61, 32, 31, 53, create an expression that equals 159. You may use +, -, *(...TRUNCATED) | "{\"numbers\": [74, 74, 61, 32, 31, 53], \"target\": 159, \"operators_used\": [\"-\", \"+\", \"+\", (...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 159 using each of the numbers 74, 74, 61, 32, 31, 53 exa(...TRUNCATED) | [
25155
] | [
31
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 159 using each of the numbers 74, 74, 61, 32, 31, 53 exa(...TRUNCATED) | iterative | [
"answer_correct"
] | [
31
] | true | [[["53 × (74 ÷ 74 - 61 + 32 + 31)"],["53 × (74 ÷ 74 - 61 + 32 + 31)"],["53 × (74 ÷ 74 - 61 + 3(...TRUNCATED) | [[[true],[true],[true],[true,true],[true],[true],[true,true],[true],[true,true],[true],[true,true],[(...TRUNCATED) | [[[{"answer_block":"53 * (74 / 74 - 61 + 32 + 31)","error":null,"final_answer":159.0,"is_correct":tr(...TRUNCATED) |
"Using the numbers 25, 69, 3, 18, 55, 1, create an expression that equals 612. You may use +, -, *, (...TRUNCATED) | "{\"numbers\": [25, 69, 3, 18, 55, 1], \"target\": 612, \"operators_used\": [\"*\", \"/\", \"-\", \"(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 612 using the numbers 25, 69, 3, 18, 55, 1. Each number (...TRUNCATED) | [
33111
] | [
28
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 612 using the numbers 25, 69, 3, 18, 55, 1. Each number (...TRUNCATED) | iterative | [
null
] | [
null
] | true | [[["(69 - 1) * (55 - 25 - 3 - 18)","(69 - 1) * (55 - 25 - 3 - 18)"],["(69 - 1) * (55 - 25 - 3 - 18)"(...TRUNCATED) | [[[true,true],[true],[true],[true,true],[true],[true],[true,true],[true,true],[true],[true,true],[tr(...TRUNCATED) | [[[{"answer_block":"(69 - 1) * (55 - 25 - 3 - 18)","error":null,"final_answer":612.0,"is_correct":tr(...TRUNCATED) |
"Using the numbers 43, 1, 17, 82, 34, 21, create an expression that equals 868. You may use +, -, *,(...TRUNCATED) | "{\"numbers\": [43, 1, 17, 82, 34, 21], \"target\": 868, \"operators_used\": [\"/\", \"*\", \"+\", \(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 868 using the numbers 43, 1, 17, 82, 34, 21. E(...TRUNCATED) | [
30327
] | [
31
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 868 using the numbers 43, 1, 17, 82, 34, 21. E(...TRUNCATED) | iterative | [
"answer_correct"
] | [
31
] | true | [[["82 * ((43 + 1) / (21 - 17)) - 34"],["82 * ((43 + 1) / (21 - 17)) - 34"],["82 * ((43 + 1) / (21 -(...TRUNCATED) | [[[true],[true],[true,true],[true],[true],[true,true],[true],[true,true],[true],[true,true],[true],[(...TRUNCATED) | [[[{"answer_block":"82 * ((43 + 1) / (21 - 17)) - 34","error":null,"final_answer":868.0,"is_correct"(...TRUNCATED) |
"Using the numbers 95, 43, 53, 86, 58, 32, create an expression that equals 103. You may use +, -, *(...TRUNCATED) | "{\"numbers\": [95, 43, 53, 86, 58, 32], \"target\": 103, \"operators_used\": [\"*\", \"-\", \"+\", (...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 103 using each of the numbers 95, 43, 53, 86, 58, 32 at (...TRUNCATED) | [
31980
] | [
31
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 103 using each of the numbers 95, 43, 53, 86, 58, 32 at (...TRUNCATED) | iterative | [
"answer_correct"
] | [
31
] | true | [[["43 + 32 + 86 - 58"],["43 + 32 + 86 - 58"],["86 + 43 - (58 - 32)","86 + 43 - (58 - 32)"],["86 + 4(...TRUNCATED) | [[[true],[true],[true,true],[true],[true],[true],[true],[true,true],[true],[true],[true,true],[true](...TRUNCATED) | [[[{"answer_block":"43 + 32 + 86 - 58","error":null,"final_answer":103.0,"is_correct":true}],[{"answ(...TRUNCATED) |
"Using the numbers 87, 74, 76, 3, 17, 1, create an expression that equals 383. You may use +, -, *, (...TRUNCATED) | "{\"numbers\": [87, 74, 76, 3, 17, 1], \"target\": 383, \"operators_used\": [\"*\", \"+\", \"-\", \"(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 383 using the numbers 87, 74, 76, 3, 17, 1. Each number (...TRUNCATED) | [
22089
] | [
31
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 383 using the numbers 87, 74, 76, 3, 17, 1. Each number (...TRUNCATED) | iterative | [
"answer_correct"
] | [
31
] | true | [[["(87 + 76) * (3 - 1) + (74 - 17)"],["((87 + 76) * (3 - 1)) + (74 - 17)","((87 + 76) * (3 - 1)) + (...TRUNCATED) | [[[true],[true,true],[true],[true],[true],[true],[true],[true],[true],[true],[true,true],[true],[tru(...TRUNCATED) | [[[{"answer_block":"(87 + 76) * (3 - 1) + (74 - 17)","error":null,"final_answer":383.0,"is_correct":(...TRUNCATED) |
"Using the numbers 89, 1, 84, 42, 43, 87, create an expression that equals 308. You may use +, -, *,(...TRUNCATED) | "{\"numbers\": [89, 1, 84, 42, 43, 87], \"target\": 308, \"operators_used\": [\"/\", \"*\", \"/\", \(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 308 using the numbers 89, 1, 84, 42, 43, 87. E(...TRUNCATED) | [
32878
] | [
25
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 308 using the numbers 89, 1, 84, 42, 43, 87. E(...TRUNCATED) | iterative | [
null
] | [
null
] | true | [[["((89 + 87) * (84 / 42)) - (43 + 1)"],["(89 + 87) * (84 / 42) - 43 - 1"],["(89 + 87) * (84 / 42) (...TRUNCATED) | [[[true],[true],[true,true],[true],[true],[true,true],[true],[true,true],[true],[true,true],[true,tr(...TRUNCATED) | [[[{"answer_block":"((89 + 87) * (84 / 42)) - (43 + 1)","error":null,"final_answer":308.0,"is_correc(...TRUNCATED) |
"Using the numbers 19, 1, 29, 87, 8, 87, create an expression that equals 145. You may use +, -, *, (...TRUNCATED) | "{\"numbers\": [19, 1, 29, 87, 8, 87], \"target\": 145, \"operators_used\": [\"/\", \"*\", \"-\", \"(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 145 using the numbers 19, 1, 29, 87, 8, 87. Wa(...TRUNCATED) | [
23168
] | [
31
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 145 using the numbers 19, 1, 29, 87, 8, 87. Wa(...TRUNCATED) | iterative | [
"answer_correct"
] | [
31
] | true | [[["87 + 87 - 29"],["87 + 87 - 29","87 + 87 - 29"],["87 + 87 - 29"],["87 + 87 - 29","87 + 87 - 29"],(...TRUNCATED) | [[[true],[true,true],[true],[true,true],[true],[true],[true,true],[true],[true],[true,true],[true],[(...TRUNCATED) | [[[{"answer_block":"87 + 87 - 29","error":null,"final_answer":145.0,"is_correct":true}],[{"answer_bl(...TRUNCATED) |
"Using the numbers 75, 49, 1, 48, 14, 87, create an expression that equals 949. You may use +, -, *,(...TRUNCATED) | "{\"numbers\": [75, 49, 1, 48, 14, 87], \"target\": 949, \"operators_used\": [\"-\", \"/\", \"+\", \(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 949 using the numbers 75, 49, 1, 48, 14, 87. E(...TRUNCATED) | [
33046
] | [
17
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's try to figure out how to make 949 using the numbers 75, 49, 1, 48, 14, 87. E(...TRUNCATED) | iterative | [
null
] | [
null
] | true | [[["(75 - 1) * 14 - 87"],["(75 - 1) * 14 - 87","(75 - 1) * 14 - 87"],["(75 - 1) * 14 - 87","(75 - 1)(...TRUNCATED) | [[[true],[true,true],[true,true,true],[true],[true,true,true],[true,true],[true,true],[true],[true],(...TRUNCATED) | [[[{"answer_block":"(75 - 1) * 14 - 87","error":null,"final_answer":949.0,"is_correct":true}],[{"ans(...TRUNCATED) |
"Using the numbers 84, 53, 21, 18, 6, 5, create an expression that equals 231. You may use +, -, *, (...TRUNCATED) | "{\"numbers\": [84, 53, 21, 18, 6, 5], \"target\": 231, \"operators_used\": [\"*\", \"/\", \"+\", \"(...TRUNCATED) | countdown | [{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 231 using the numbers 84, 53, 21, 18, 6, 5. Each number (...TRUNCATED) | [
29361
] | [
32
] | [["This is actually wrong, try again.","This is actually wrong, try again.","This is actually wrong,(...TRUNCATED) | [["<think>\nOkay, let's see. I need to make 231 using the numbers 84, 53, 21, 18, 6, 5. Each number (...TRUNCATED) | iterative | [
"answer_correct"
] | [
32
] | true | [[["(84 * 18 / 6) - 21"],["21 * (5 + 6)","21 * (5 + 6)"],["21 * (5 + 6)"],["21 * (5 + 6)","21 * (5 +(...TRUNCATED) | [[[true],[true,true],[true],[true,true],[true],[true,true],[false],[false,false],[false],[false],[fa(...TRUNCATED) | [[[{"answer_block":"(84 * 18 / 6) - 21","error":null,"final_answer":231.0,"is_correct":true}],[{"ans(...TRUNCATED) |
End of preview. Expand in Data Studio
TIW-Qwen3-4B-Thinking-2507-cd6arg-iterative
Countdown reasoning with iterative strategy to 32768 tokens
Dataset Info
- Rows: 100
- Columns: 16
Columns
| Column | Type | Description |
|---|---|---|
| question | Value('string') | The countdown problem statement |
| metadata | Value('string') | JSON with target number, available numbers, and solution |
| task_source | Value('string') | Task type identifier (countdown) |
| formatted_prompt | List({'content': Value('string'), 'role': Value('string')}) | Initial prompt messages sent to the model |
| responses_by_sample | List(List(Value('string'))) | Full model response [sample_idx][round_idx] = response string |
| tokens_by_sample | List(Value('int64')) | Total token count for each sample (based on continuation_text) |
| num_rounds_by_sample | List(Value('int64')) | Number of continuation rounds for each sample |
| continuation_prompts_by_sample | List(List(Value('string'))) | Continuation prompts used [sample_idx][round_idx] = prompt string |
| continuation_text_by_sample | List(List(Value('string'))) | Truncated text used for continuation [sample_idx][round_idx]. For budget_forcing: truncated at first tag. For other strategies: same as responses_by_sample. |
| strategy_type | Value('string') | Prompting strategy: 'iterative', 'budget_forcing', or 'random_noise' |
| termination_reason_by_sample | List(Value('string')) | Why each sample terminated: 'max_rounds_reached', 'budget_reached', or 'answer_correct' |
| correct_round_by_sample | List(Value('int64')) | Round number where correct answer was first found (None if never correct) |
| smart_termination_enabled | Value('bool') | Whether smart termination was enabled for this run |
| answers_by_sample | List(List(List(Value('string')))) | Extracted answer spans [sample][round][answer_idx] |
| evals_by_sample | List(List(List(Value('bool')))) | Correctness evaluations [sample][round][answer_idx] |
| eval_metadata_by_sample | List(List(List({'answer_block': Value('string'), 'error': Value('string'), 'final_answer': Value('float64'), 'is_correct': Value('bool')}))) | Evaluation details [sample][round][answer_idx] |
Generation Parameters
{
"script_name": "stage_01_generate_reasoning.py",
"model": "hosted_vllm/Qwen/Qwen3-4B-Thinking-2507<api_base>http://10.32.36.199:9080/v1",
"hyperparameters": {
"temperature": 0.7,
"max_tokens_per_round": 32768,
"target_total_tokens": 32768,
"samples": 1,
"strategy": "iterative",
"max_rounds": 50
},
"input_datasets": [],
"description": "Countdown reasoning with iterative strategy to 32768 tokens",
"experiment": "degen_test_1_countdown",
"strategy": "iterative",
"arg_count": 6,
"seed": 42,
"smart_termination": true,
"max_rounds": 50,
"status": "completed",
"completed_at": "2025-12-22T03:24:06.941130"
}
Experiment Documentation
For complete experiment details, see https://github.com/TAUR-Lab/Reasoning-Horizons/tree/main/experiments/degen_test_1_countdown
Usage
from datasets import load_dataset
dataset = load_dataset("reasoning-degeneration-dev/TIW-Qwen3-4B-Thinking-2507-cd6arg-iterative", split="train")
print(f"Loaded {len(dataset)} rows")
This dataset is tracked in reasoning-degeneration-dev/PROJECT-MANIFEST
- Downloads last month
- 8