Dataset Viewer
Auto-converted to Parquet Duplicate
question
stringlengths
152
158
metadata
stringlengths
160
172
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
retry_count_by_sample
listlengths
1
1
answers_by_sample
listlengths
1
1
evals_by_sample
listlengths
1
1
eval_metadata_by_sample
listlengths
1
1
"Using ALL of the numbers 12, 4, 7, 1, 2, 5, 21, 88, create an expression that equals 109. You must (...TRUNCATED)
"{\"numbers\": [12, 4, 7, 1, 2, 5, 21, 88], \"target\": 109, \"operators_used\": [\"/\", \"+\", \"/\(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**12, 4, 7, 1, 2, 5, 21, 88** \nWe must use **each number exactly o(...TRUNCATED)
[ 61345 ]
[ 1 ]
[ [] ]
[["We are given the numbers: \n**12, 4, 7, 1, 2, 5, 21, 88** \nWe must use **each number exactly o(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["88 + 21 + (12 - 12) + (7 - 7) + (5 - 5) + (4 - 4) + (1 - 1) + (2 - 2)","88 + 21 + (12 - 12) + (7(...TRUNCATED)
[[[false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,f(...TRUNCATED)
[[[{"answer_block":"88 + 21 + (12 - 12) + (7 - 7) + (5 - 5) + (4 - 4) + (1 - 1) + (2 - 2)","error":"(...TRUNCATED)
"Using ALL of the numbers 25, 69, 58, 14, 55, 24, 36, 60, create an expression that equals 176. You (...TRUNCATED)
"{\"numbers\": [25, 69, 58, 14, 55, 24, 36, 60], \"target\": 176, \"operators_used\": [\"-\", \"+\",(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**25, 69, 58, 14, 55, 24, 36, 60** \nWe must use **all of them exac(...TRUNCATED)
[ 74039 ]
[ 7 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**25, 69, 58, 14, 55, 24, 36, 60** \nWe must use **all of them exac(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["55 + 24 + 60 + 36 + (69 - 58) / (25 - 14)"],["55 + 24 + 60 + 36 + (69 - 58) / (25 - 14)"],["55 +(...TRUNCATED)
[ [ [ true ], [ true ], [ true ], [ true ], [ true ], [ true ], [ true ] ] ]
[[[{"answer_block":"55 + 24 + 60 + 36 + (69 - 58) / (25 - 14)","error":null,"final_answer":176.0,"is(...TRUNCATED)
"Using ALL of the numbers 55, 36, 6, 3, 14, 99, 17, 82, create an expression that equals 245. You mu(...TRUNCATED)
"{\"numbers\": [55, 36, 6, 3, 14, 99, 17, 82], \"target\": 245, \"operators_used\": [\"*\", \"-\", \(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**55, 36, 6, 3, 14, 99, 17, 82** \nWe must use **each number exactl(...TRUNCATED)
[ 76480 ]
[ 13 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**55, 36, 6, 3, 14, 99, 17, 82** \nWe must use **each number exactl(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["55 + 36 + 6 + 3 + 14 + 99 + 17 + 82 - (82 - (17 - (6 / 3)))"],["55 + 36 + 6 + 3 + 14 + 99 + 17 +(...TRUNCATED)
[[[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[f(...TRUNCATED)
[[[{"answer_block":"55 + 36 + 6 + 3 + 14 + 99 + 17 + 82 - (82 - (17 - (6 / 3)))","error":"Expression(...TRUNCATED)
"Using ALL of the numbers 87, 87, 29, 2, 19, 4, 1, 1, create an expression that equals 281. You must(...TRUNCATED)
"{\"numbers\": [87, 87, 29, 2, 19, 4, 1, 1], \"target\": 281, \"operators_used\": [\"/\", \"+\", \"/(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**87, 87, 29, 2, 19, 4, 1, 1** \nWe must use **each number exactly (...TRUNCATED)
[ 70302 ]
[ 19 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**87, 87, 29, 2, 19, 4, 1, 1** \nWe must use **each number exactly (...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["87 × (4 − 1) + (29 − 19) × 2 + (87 − 87) + (1 − 1)"],["87 × (4 - 1) + (29 - 19) × 2 (...TRUNCATED)
[[[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[f(...TRUNCATED)
[[[{"answer_block":"87 * (4 − 1) + (29 − 19) * 2 + (87 − 87) + (1 − 1)","error":"Could not e(...TRUNCATED)
"Using ALL of the numbers 56, 63, 7, 50, 44, 1, 87, 52, create an expression that equals 132. You mu(...TRUNCATED)
"{\"numbers\": [56, 63, 7, 50, 44, 1, 87, 52], \"target\": 132, \"operators_used\": [\"-\", \"/\", \(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**56, 63, 7, 50, 44, 1, 87, 52** \nWe must use **each number exactl(...TRUNCATED)
[ 69254 ]
[ 21 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**56, 63, 7, 50, 44, 1, 87, 52** \nWe must use **each number exactl(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["63 - 56 - 7 + 87 + 52 + 44 - 50 - 1"],["63 - 56 - 7 + 87 + 52 + 44 - 50 - 1"],["63 - 56 - 7 + 87(...TRUNCATED)
[[[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],(...TRUNCATED)
[[[{"answer_block":"63 - 56 - 7 + 87 + 52 + 44 - 50 - 1","error":null,"final_answer":132.0,"is_corre(...TRUNCATED)
"Using ALL of the numbers 7, 1, 49, 7, 28, 59, 3, 44, create an expression that equals 521. You must(...TRUNCATED)
"{\"numbers\": [7, 1, 49, 7, 28, 59, 3, 44], \"target\": 521, \"operators_used\": [\"/\", \"*\", \"/(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**7, 1, 49, 7, 28, 59, 3, 44** \nWe must use **all of them exactly (...TRUNCATED)
[ 59415 ]
[ 1 ]
[ [] ]
[["We are given the numbers: \n**7, 1, 49, 7, 28, 59, 3, 44** \nWe must use **all of them exactly (...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["59 × ((28 / 7) + (7 - 3)) + 49 + (1 - 1)","59 * ((28 / 7) + (7 - 3)) + 49 + (44 - 44) + (1 - 1)(...TRUNCATED)
[[[false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,f(...TRUNCATED)
[[[{"answer_block":"59 * ((28 / 7) + (7 - 3)) + 49 + (1 - 1)","error":"Expression uses an unavailabl(...TRUNCATED)
"Using ALL of the numbers 97, 4, 93, 26, 3, 24, 20, 31, create an expression that equals 620. You mu(...TRUNCATED)
"{\"numbers\": [97, 4, 93, 26, 3, 24, 20, 31], \"target\": 620, \"operators_used\": [\"-\", \"/\", \(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers:\n\n**97, 4, 93, 26, 3, 24, 20, 31**\n\nWe must **use each number exactl(...TRUNCATED)
[ 58109 ]
[ 1 ]
[ [] ]
[["We are given the numbers:\n\n**97, 4, 93, 26, 3, 24, 20, 31**\n\nWe must **use each number exactl(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["(20 * 31) + (97 - 97) + (26 - 26) + (24 - 24) + (4 - 4) + (3 - 3)","(24 * 26) - 4 + (31 * 3 - 93(...TRUNCATED)
[[[false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,f(...TRUNCATED)
[[[{"answer_block":"(20 * 31) + (97 - 97) + (26 - 26) + (24 - 24) + (4 - 4) + (3 - 3)","error":"Expr(...TRUNCATED)
"Using ALL of the numbers 45, 69, 55, 40, 1, 9, 8, 83, create an expression that equals 913. You mus(...TRUNCATED)
"{\"numbers\": [45, 69, 55, 40, 1, 9, 8, 83], \"target\": 913, \"operators_used\": [\"*\", \"+\", \"(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers:\n\n**45, 69, 55, 40, 1, 9, 8, 83**\n\nWe must use **each number exactly(...TRUNCATED)
[ 70077 ]
[ 18 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers:\n\n**45, 69, 55, 40, 1, 9, 8, 83**\n\nWe must use **each number exactly(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["(69 + 40 + 1 - 55 - 45 + 9 - 8) × 83 = 913"],["(69 + 40 + 1 - 55 - 45 + 9 - 8) × 83"],["(69 + (...TRUNCATED)
[[[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],[true],(...TRUNCATED)
[[[{"answer_block":"(69 + 40 + 1 - 55 - 45 + 9 - 8) * 83 = 913","error":null,"final_answer":913.0,"i(...TRUNCATED)
"Using ALL of the numbers 72, 3, 67, 1, 43, 1, 93, 53, create an expression that equals 145. You mus(...TRUNCATED)
"{\"numbers\": [72, 3, 67, 1, 43, 1, 93, 53], \"target\": 145, \"operators_used\": [\"/\", \"-\", \"(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**72, 3, 67, 1, 43, 1, 93, 53** \nWe must use **all of them exactly(...TRUNCATED)
[ 75769 ]
[ 22 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**72, 3, 67, 1, 43, 1, 93, 53** \nWe must use **all of them exactly(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["(93 + 53 - 1) + (1 - 1) + (72 / 3 - (67 - 43))"],["(93 + 53 - 1) + (1 - 1) + (72 / 3 - (67 - 43)(...TRUNCATED)
[[[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[f(...TRUNCATED)
[[[{"answer_block":"(93 + 53 - 1) + (1 - 1) + (72 / 3 - (67 - 43))","error":"Expression uses an unav(...TRUNCATED)
"Using ALL of the numbers 13, 1, 13, 68, 2, 2, 93, 62, create an expression that equals 102. You mus(...TRUNCATED)
"{\"numbers\": [13, 1, 13, 68, 2, 2, 93, 62], \"target\": 102, \"operators_used\": [\"/\", \"-\", \"(...TRUNCATED)
countdown
[{"content":"You are a helpful assistant that solves math problems by reasoning step-by-step. Show y(...TRUNCATED)
[["We are given the numbers: \n**13, 1, 13, 68, 2, 2, 93, 62** \nWe must use **all of them exactly(...TRUNCATED)
[ 74416 ]
[ 18 ]
[[" Wait… I’ll continue exploring alternative solution strategies. Let’s try something new. ",(...TRUNCATED)
[["We are given the numbers: \n**13, 1, 13, 68, 2, 2, 93, 62** \nWe must use **all of them exactly(...TRUNCATED)
budget_forcing
[ null ]
[ null ]
false
[ 0 ]
[[["(93 + 68 - 62 + 13 - 13 + 1 + 2) * (2 / 2)"],["(93 + 68 - 62 + 13 - 13 + 1 + 2) * (2 / 2)"],["(9(...TRUNCATED)
[[[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[false],[f(...TRUNCATED)
[[[{"answer_block":"(93 + 68 - 62 + 13 - 13 + 1 + 2) * (2 / 2)","error":"Expression uses an unavaila(...TRUNCATED)
End of preview. Expand in Data Studio

EAS-qwen3-4B-Instruct-2507-cd8arg-budget_forcing

Countdown reasoning with budget_forcing strategy to 32768 tokens

Dataset Info

  • Rows: 100
  • Columns: 17

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('null')) 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
retry_count_by_sample List(Value('int64')) Total retry attempts per sample due to empty or too-short responses (0 if no retries needed)
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-Instruct-2507<api_base>http://10.32.37.15:10080/v1",
  "hyperparameters": {
    "temperature": 0.7,
    "max_tokens_per_round": 32768,
    "target_total_tokens": 32768,
    "samples": 1,
    "strategy": "budget_forcing",
    "max_rounds": null,
    "min_tokens": 10,
    "max_retries_per_round": -1
  },
  "input_datasets": [],
  "description": "Countdown reasoning with budget_forcing strategy to 32768 tokens",
  "experiment": "degen_test_1_countdown",
  "strategy": "budget_forcing",
  "arg_count": 8,
  "seed": 42,
  "smart_termination": false,
  "max_rounds": null,
  "min_tokens": 10,
  "inference_backend": "inference_engine",
  "status": "completed",
  "completed_at": "2025-12-26T07:51:16.097656"
}

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/EAS-qwen3-4B-Instruct-2507-cd8arg-budget_forcing", split="train")
print(f"Loaded {len(dataset)} rows")

This dataset is tracked in reasoning-degeneration-dev/PROJECT-MANIFEST

Downloads last month
10