Dataset Viewer
Auto-converted to Parquet Duplicate
question
stringlengths
973
978
answer
stringlengths
5
14
task_config
stringclasses
1 value
task_source
stringclasses
1 value
prompt
listlengths
1
1
model_responses
listlengths
0
0
model_responses__eval_is_correct
listlengths
0
0
all_other_columns
stringlengths
171
185
original_split
stringclasses
1 value
metadata
stringlengths
113
127
model_responses__best_of_n_atags
listlengths
4
4
model_responses__best_of_n_atags__finish_reason_length_flags
listlengths
4
4
model_responses__best_of_n_atags__length_partial_responses
listlengths
4
4
prompt__best_of_n_atags__metadata
dict
model_responses__best_of_n_atags__metadata
dict
model_responses__best_of_n_atags__eval_is_correct
listlengths
4
4
model_responses__best_of_n_atags__eval_extracted_answers
listlengths
4
4
model_responses__best_of_n_atags__eval_extraction_metadata
stringlengths
374
6.8k
model_responses__best_of_n_atags__eval_evaluation_metadata
stringlengths
364
6.7k
model_responses__best_of_n_atags__internal_answers__eval_is_correct
listlengths
4
4
model_responses__best_of_n_atags__internal_answers__eval_extracted_answers
listlengths
4
4
model_responses__best_of_n_atags__internal_answers__eval_extraction_metadata
stringlengths
1.45k
58.8k
model_responses__best_of_n_atags__internal_answers__eval_evaluation_metadata
stringlengths
803
36.9k
model_responses__best_of_n_atags__metrics
dict
eval_date
stringdate
2025-11-10 00:00:00
2025-11-10 00:00:00
split
stringclasses
1 value
revision_name
stringclasses
1 value
model_path
stringclasses
1 value
checkpoint_step
int64
0
0
stage_name
stringclasses
1 value
stage_number
int64
1
1
timestamp
stringdate
2025-11-10 02:50:48
2025-11-10 02:50:48
eval_repo_id
stringclasses
1 value
Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>. # Problem Using the numbers in the list [96, 97, 68], create an equation that equals 125. You can use basic arithmetic operations (+, -, *, /) and each number can o...
97 + 96 - 68
countdown_3arg
countdown
[ { "content": "Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>.\n\n# Problem\nUsing the numbers in the list [96, 97, 68], create an equation that equals 125. You can use basic arithmetic operations (+, -, *,...
[]
[]
{"metadata": "{\"source_dataset\": \"countdown\", \"source_index\": 5000, \"numbers\": [96, 97, 68], \"target\": 125, \"expression\": \"97 + 96 - 68\"}", "task_source": "countdown"}
test
{"source_dataset": "countdown", "source_index": 5000, "numbers": [96, 97, 68], "target": 125, "expression": "97 + 96 - 68"}
[ "<think>\n<sample>\nRewritten Question:\n\nCreate an equation using the numbers from the set {96, 97, 68} which equals exactly 125.\n\nStep 1: Start with one of the given numbers.\n - Let's start with 96.\n\nStep 2: Add another number to form a product if possible.\n - We need to add either 97 or 68 to get clos...
[ false, false, false, false ]
[ "", "", "", "" ]
{ "api_url": "http://10.32.37.26:9080/v1", "backend": "litellm", "chat_template_applied": true, "generation_params": { "chat_template_applied": true, "max_tokens": 16384, "n": 1, "repetition_penalty": 1.1, "temperature": 0.7, "top_k": 20, "top_p": 0.8 }, "model_name": "hosted_vll...
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ true, false, true, true ]
[ "(96 + 97 - 68)", "((96 × 97) ÷ 68)", "(96 + 97) - 68 = 125", "96 + 97 - 68" ]
[{"extraction_type": "final", "total_spans_found": 4, "non_overlapping_spans": 4, "span_start": 3415, "span_end": 3431, "original_span": "(96 + 97 - 68)"}, {"extraction_type": "final", "total_spans_found": 3, "non_overlapping_spans": 3, "span_start": 2127, "span_end": 2145, "original_span": "((96 \u00d7 97) \u00f7 68)"...
[{"is_correct": true, "error": null, "answer_block": "(96 + 97 - 68)", "final_answer": 125.0}, {"is_correct": false, "error": "Could not evaluate expression", "answer_block": "((96 \u00d7 97) \u00f7 68)", "final_answer": null}, {"is_correct": true, "error": null, "answer_block": "(96 + 97) - 68 = 125", "final_answer": ...
[ [ false, true, true, true ], [ true, false, false ], [ false, false, true, true ], [ false, true, true ] ]
[ [ "\\( 96 + 97 - 68 = 125 \\) or \\( 96 + 68 - 68 = 125 \\)", "(96 + 97 - 68)", "(96 + 97 - 68)", "(96 + 97 - 68)" ], [ "((96 + 97) - 68)", "((96 × 97) ÷ 68)", "((96 × 97) ÷ 68)" ], [ "\\( 96 + 97 - 68 = 125 \\)", "((96 * 97) / 68) = 125", "(96 + 97) - 68 = 125", ...
[[{"pattern_used": "answer tag", "confidence": 1.0, "position": 1024, "span_text": "\\( 96 + 97 - 68 = 125 \\) or \\( 96 + 68 - 68 = 125 \\)", "extraction_type": "internal", "span_start": 1024, "span_end": 1078, "original_span": "\\( 96 + 97 - 68 = 125 \\) or \\( 96 + 68 - 68 = 125 \\)"}, {"pattern_used": "answer tag",...
[[{"is_correct": false, "error": "Could not evaluate expression", "answer_block": "\\( 96 + 97 - 68 = 125 \\) or \\( 96 + 68 - 68 = 125 \\)", "final_answer": null}, {"is_correct": true, "error": null, "answer_block": "(96 + 97 - 68)", "final_answer": 125.0}, {"is_correct": true, "error": null, "answer_block": "(96 + 97...
{ "flips_by": [ 0, 1, 0, 0 ], "flips_total": 1, "num_correct": 3, "pass_at_n": 1, "percent_correct": 0.75, "skill_count": { "answer_revision": [ 0, 0, 0, 0 ], "best_of_n": [ 0, 0, 1, 0 ], "reflect_close": [ 2, ...
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472418
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>. # Problem Using the numbers in the list [21, 80, 46], create an equation that equals 105. You can use basic arithmetic operations (+, -, *, /) and each number can o...
46 + 80 - 21
countdown_3arg
countdown
[ { "content": "Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>.\n\n# Problem\nUsing the numbers in the list [21, 80, 46], create an equation that equals 105. You can use basic arithmetic operations (+, -, *,...
[]
[]
{"metadata": "{\"source_dataset\": \"countdown\", \"source_index\": 5001, \"numbers\": [21, 80, 46], \"target\": 105, \"expression\": \"46 + 80 - 21\"}", "task_source": "countdown"}
test
{"source_dataset": "countdown", "source_index": 5001, "numbers": [21, 80, 46], "target": 105, "expression": "46 + 80 - 21"}
[ "<think>\n<sample>\nFirst, I'll try to understand the problem better by writing out a plan and go really deep into detail about how I should solve this.\nThe goal is to create an equation using the numbers provided (21, 80, 46) which will equal 105. Each number must be used exactly once.\nI will start by trying dif...
[ false, false, false, false ]
[ "", "", "", "" ]
{ "api_url": "http://10.32.37.26:9080/v1", "backend": "litellm", "chat_template_applied": true, "generation_params": { "chat_template_applied": true, "max_tokens": 16384, "n": 1, "repetition_penalty": 1.1, "temperature": 0.7, "top_k": 20, "top_p": 0.8 }, "model_name": "hosted_vll...
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ true, false, false, false ]
[ "(80 - 21) + 46 = 105", "\\(80 - 21 + 46 = 105\\)", "(80 - 21) * 46 / 26", "(80 - 21) * 46 = 105" ]
[{"extraction_type": "final", "total_spans_found": 8, "non_overlapping_spans": 8, "span_start": 9843, "span_end": 9865, "original_span": "(80 - 21) + 46 = 105"}, {"extraction_type": "final", "total_spans_found": 3, "non_overlapping_spans": 3, "span_start": 2826, "span_end": 2850, "original_span": "\\(80 - 21 + 46 = 105...
[{"is_correct": true, "error": null, "answer_block": "(80 - 21) + 46 = 105", "final_answer": 105.0}, {"is_correct": false, "error": "Could not evaluate expression", "answer_block": "\\(80 - 21 + 46 = 105\\)", "final_answer": null}, {"is_correct": false, "error": "Expression does not equal stated result", "answer_block"...
[ [ false, false, false, false, false, false, true, true ], [ true, false, false ], [ false, false, false ], [ false, false ] ]
[ [ "((80 * 46) - 21) / 21 = 105", "\\( 80 \\times 2 + 21 - 46 \\)", "It is impossible to create an equation using the numbers 21, 80, and 46 that equals 105.", "\\( (101 - 46) * 2 = 105 \\)", "(21 * 5) - 80 + 46 = 105", "(21 + 80) * 46 = 105", "(80 - 21) + 46 = 105", "(80 - 21) + 46 =...
[[{"pattern_used": "answer tag", "confidence": 1.0, "position": 1003, "span_text": "((80 * 46) - 21) / 21 = 105", "extraction_type": "internal", "span_start": 1003, "span_end": 1030, "original_span": "((80 * 46) - 21) / 21 = 105"}, {"pattern_used": "answer tag", "confidence": 1.0, "position": 3452, "span_text": "\\( 80...
[[{"is_correct": false, "error": "Expression does not equal stated result", "answer_block": "((80 * 46) - 21) / 21 = 105", "final_answer": 105.0}, {"is_correct": false, "error": "Could not evaluate expression", "answer_block": "\\( 80 \\times 2 + 21 - 46 \\)", "final_answer": null}, {"is_correct": false, "error": "Coul...
{ "flips_by": [ 0, 1, 0, 0 ], "flips_total": 1, "num_correct": 1, "pass_at_n": 1, "percent_correct": 0.25, "skill_count": { "answer_revision": [ 0, 0, 0, 0 ], "best_of_n": [ 1, 0, 0, 0 ], "reflect_close": [ 7, ...
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472429
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>. # Problem Using the numbers in the list [6, 76, 96], create an equation that equals 500. You can use basic arithmetic operations (+, -, *, /) and each number can on...
6*96 - 76
countdown_3arg
countdown
[ { "content": "Answer the following problem. Explain your reasoning step by step. When you are finished, give your answer in this format: <answer>(your answer)</answer>.\n\n# Problem\nUsing the numbers in the list [6, 76, 96], create an equation that equals 500. You can use basic arithmetic operations (+, -, *, ...
[]
[]
{"metadata": "{\"source_dataset\": \"countdown\", \"source_index\": 5002, \"numbers\": [6, 76, 96], \"target\": 500, \"expression\": \"6*96 - 76\"}", "task_source": "countdown"}
test
{"source_dataset": "countdown", "source_index": 5002, "numbers": [6, 76, 96], "target": 500, "expression": "6*96 - 76"}
[ "<think>\n<sample>\nFirst, I'll try to understand the problem better by writing out a plan and go really deep into detail about how I should solve this.\n- The goal is to create an equation using the numbers provided (6, 76, 96) which will equal 500.\n- Each number must be used exactly once.\n- We can use addition,...
[ false, false, false, false ]
[ "", "", "", "" ]
{ "api_url": "http://10.32.37.26:9080/v1", "backend": "litellm", "chat_template_applied": true, "generation_params": { "chat_template_applied": true, "max_tokens": 16384, "n": 1, "repetition_penalty": 1.1, "temperature": 0.7, "top_k": 20, "top_p": 0.8 }, "model_name": "hosted_vll...
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, false, false, false ]
[ "Alternatively,\n\nStep 1: (6 * 76) - 96 = 456\n\nThis uses all three numbers exactly once and achieves the target of 500 when added together: 456 + 44 = 500.", "\\( 500 - (76 + 96 + 6) \\)", "((76 + 96) - 6) * 6", "\\( (76 \\times 96) / 6 \\approx 1229.333... \\)" ]
[{"extraction_type": "final", "total_spans_found": 4, "non_overlapping_spans": 4, "span_start": 7353, "span_end": 7508, "original_span": "Alternatively,\n\nStep 1: (6 * 76) - 96 = 456\n\nThis uses all three numbers exactly once and achieves the target of 500 when added together: 456 + 44 = 500."}, {"extraction_type": "...
[{"is_correct": false, "error": "Incorrect arithmetic in line: Step 1: (6 * 76) - 96 = 456", "answer_block": "Alternatively,\n\nStep 1: (6 * 76) - 96 = 456\n\nThis uses all three numbers exactly once and achieves the target of 500 when added together: 456 + 44 = 500.", "final_answer": 456.0}, {"is_correct": false, "err...
[ [ false, false, false, false ], [ false, false, false, false, false ], [ false, false, false, false, false, false, false, false ], [ false, false, false, false ] ]
[ [ "(76 + 96 - 6*7 - 6*5 - 6*4 - 6*3 - 6*2 - 6*1 - 6*0)", "(\\frac{500}{10}) \\times 10", "((76 + 96) - 500) / 6", "Alternatively,\n\nStep 1: (6 * 76) - 96 = 456\n\nThis uses all three numbers exactly once and achieves the target of 500 when added together: 456 + 44 = 500." ], [ "(6 * 76 - 4)...
[[{"pattern_used": "answer tag", "confidence": 1.0, "position": 3097, "span_text": "(76 + 96 - 6*7 - 6*5 - 6*4 - 6*3 - 6*2 - 6*1 - 6*0)", "extraction_type": "internal", "span_start": 3097, "span_end": 3150, "original_span": "(76 + 96 - 6*7 - 6*5 - 6*4 - 6*3 - 6*2 - 6*1 - 6*0)"}, {"pattern_used": "answer tag", "confiden...
[[{"is_correct": false, "error": "Expression does not equal stated result", "answer_block": "(76 + 96 - 6*7 - 6*5 - 6*4 - 6*3 - 6*2 - 6*1 - 6*0)", "final_answer": 500.0}, {"is_correct": false, "error": "Could not evaluate expression", "answer_block": "(\\frac{500}{10}) \\times 10", "final_answer": null}, {"is_correct":...
{ "flips_by": [ 0, 0, 0, 0 ], "flips_total": 0, "num_correct": 0, "pass_at_n": 0, "percent_correct": 0, "skill_count": { "answer_revision": [ 0, 0, 0, 0 ], "best_of_n": [ 0, 1, 1, 0 ], "reflect_close": [ 4, 3, ...
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472433
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
42*(12 + 3)
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5003, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5003, \"numbers\": [42, 12, 3], \"target\": 6(...TRUNCATED)
["<think>\n<sample>\nRewritten Question:\n\nCreate an equation using the numbers from the given set (...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, false, false, false ]
["\\( 42 \\times 15 = 630 \\)","(42 - 12) * 3 + 3 * 7 - 21","(42 * 12) / 3","It is not possible to a(...TRUNCATED)
"[{\"extraction_type\": \"final\", \"total_spans_found\": 4, \"non_overlapping_spans\": 4, \"span_st(...TRUNCATED)
"[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"\\\\( 42(...TRUNCATED)
[[false,false,false,false],[false,false,false,false,false,false,false,false],[false,false,false,fals(...TRUNCATED)
[["((42 * 12) / 3)","No valid equation found under standard arithmetic operations using the given nu(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 257, \"span_text\": \"((42 (...TRUNCATED)
"[[{\"is_correct\": false, \"error\": \"Expression does not equal stated result\", \"answer_block\":(...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":0,"pass_at_n":0,"percent_correct":0.0,"skill_cou(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472434
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
86 + 69 - 52
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5005, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5005, \"numbers\": [86, 69, 52], \"target\": (...TRUNCATED)
["<think>\n<sample>\nTo solve for 103 using the numbers 86, 69, and 52 from the given set, we need t(...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, true, false, true ]
[ "\\( (86 + 69) - 52 = 103 \\)", "(86 + 69 - 52)", "\\( 86 + 69 - 52 = 103 \\)", "(86 + 69) - 52" ]
"[{\"extraction_type\": \"final\", \"total_spans_found\": 2, \"non_overlapping_spans\": 2, \"span_st(...TRUNCATED)
"[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"\\\\( (8(...TRUNCATED)
[ [ false, false ], [ false, true, false, false, true, true ], [ false, false, false, false, false, false ], [ true, true ] ]
[["\\( (86 + 69) - 52 = 103 \\)","\\( (86 + 69) - 52 = 103 \\)"],["5822 - 52 - 103 = 103","((86 + 69(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 994, \"span_text\": \"\\\\((...TRUNCATED)
"[[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"\\\\( ((...TRUNCATED)
{"flips_by":[0,1,0,0],"flips_total":1,"num_correct":2,"pass_at_n":1,"percent_correct":0.5,"skill_cou(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472436
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
63*7 - 87
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5006, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5006, \"numbers\": [63, 7, 87], \"target\": 3(...TRUNCATED)
["<think>\n<sample>\nFirst, I'll try to understand the problem better by writing out a plan and go r(...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ true, true, false, false ]
[ "((63 * 7) - 87)", "(63 * 7 - 87)", "\\(441 - 87\\)", "354 - (70 + 87)" ]
"[{\"extraction_type\": \"final\", \"total_spans_found\": 2, \"non_overlapping_spans\": 2, \"span_st(...TRUNCATED)
"[{\"is_correct\": true, \"error\": null, \"answer_block\": \"((63 * 7) - 87)\", \"final_answer\": 3(...TRUNCATED)
[ [ true, true ], [ true, true ], [ false, false ], [ false, false ] ]
[["((63 * 7) - 87)","((63 * 7) - 87)"],["(63 * 7 - 87)","(63 * 7 - 87)"],["\\(441 - 87\\)","\\(441 -(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 1263, \"span_text\": \"((63(...TRUNCATED)
"[[{\"is_correct\": true, \"error\": null, \"answer_block\": \"((63 * 7) - 87)\", \"final_answer\": (...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":2,"pass_at_n":1,"percent_correct":0.5,"skill_cou(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472437
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
19*8 - 50
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5007, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5007, \"numbers\": [8, 50, 19], \"target\": 1(...TRUNCATED)
["<think>\n<sample>\nRewritten Question:\n\nCreate an equation using the numbers from the given set (...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, false, true, false ]
[ "(50 - 8) * 19 - 686", "(50 + 19) + 8 + 8 + 17", "8 * 19 - 50", "1082" ]
"[{\"extraction_type\": \"final\", \"total_spans_found\": 7, \"non_overlapping_spans\": 7, \"span_st(...TRUNCATED)
"[{\"is_correct\": false, \"error\": \"Expression does not equal stated result\", \"answer_block\": (...TRUNCATED)
[[false,false,false,false,false,false,false],[false,false,false,false,false,false,false,false,false,(...TRUNCATED)
[["No valid solution found under standard arithmetic rules with the given numbers.","\\(8 \\times 12(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 2657, \"span_text\": \"No v(...TRUNCATED)
"[[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"No vali(...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":1,"pass_at_n":1,"percent_correct":0.25,"skill_co(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472439
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
40 + 46 + 38
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5008, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5008, \"numbers\": [38, 46, 40], \"target\": (...TRUNCATED)
["<think>\n<sample>\nFirst, I'll try to understand the problem better by writing out a plan and go r(...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ true, false, true, true ]
[ "(38 + 46 + 40)", "(38 + 46 - 40) * 2 + 36", "38 + 46 + 40 = 124", "(38 + 46 + 40)" ]
"[{\"extraction_type\": \"final\", \"total_spans_found\": 2, \"non_overlapping_spans\": 2, \"span_st(...TRUNCATED)
"[{\"is_correct\": true, \"error\": null, \"answer_block\": \"(38 + 46 + 40)\", \"final_answer\": 12(...TRUNCATED)
[[true,true],[false,false,false,false,false,false,false,false,false],[false,false,true,true],[false,(...TRUNCATED)
[["(38 + 46 + 40)","(38 + 46 + 40)"],["(40 + 46 - 38)","\\(38 + 40 + 46\\)","\\( 46^2 \\)","(46 - 38(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 802, \"span_text\": \"(38 +(...TRUNCATED)
"[[{\"is_correct\": true, \"error\": null, \"answer_block\": \"(38 + 46 + 40)\", \"final_answer\": 1(...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":3,"pass_at_n":1,"percent_correct":0.75,"skill_co(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472443
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
19 + 92 + 21
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5009, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5009, \"numbers\": [92, 21, 19], \"target\": (...TRUNCATED)
["<think>\n<sample>\nFirst, I'll try to understand the problem better by writing out a plan and go r(...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, false, false, false ]
[ null, "(92 + 21 + 19) / 3", "(92 + 21 - 19 * 1.4 + 0.4)", "(92 + 21 - 19)" ]
"[{\"extraction_type\": \"final\"}, {\"extraction_type\": \"final\", \"total_spans_found\": 4, \"non(...TRUNCATED)
"[{\"evaluation_method\": \"response_processor\", \"reason\": \"no_extracted_answer\", \"extraction_(...TRUNCATED)
[ [], [ false, false, false, false ], [ false, false, false ], [ false, false, false, false ] ]
[[],["Start with 92; add 21; add 19.","\\( (92 \\times 2) - 19 + \\frac{21}{21} = 132 \\)","(92 + 21(...TRUNCATED)
"[[], [{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 786, \"span_text\": \"S(...TRUNCATED)
"[[], [{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"Sta(...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":0,"pass_at_n":0,"percent_correct":0.0,"skill_cou(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472445
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
"Answer the following problem. Explain your reasoning step by step. When you are finished, give your(...TRUNCATED)
23*(21 - 7)
countdown_3arg
countdown
[{"content":"Answer the following problem. Explain your reasoning step by step. When you are finishe(...TRUNCATED)
[]
[]
"{\"metadata\": \"{\\\"source_dataset\\\": \\\"countdown\\\", \\\"source_index\\\": 5010, \\\"number(...TRUNCATED)
test
"{\"source_dataset\": \"countdown\", \"source_index\": 5010, \"numbers\": [23, 21, 7], \"target\": 3(...TRUNCATED)
["<think>\n<sample>\nRewritten Question:\n\nCreate an equation using the numbers from the given set (...TRUNCATED)
[ false, false, false, false ]
[ "", "", "", "" ]
{"api_url":"http://10.32.37.26:9080/v1","backend":"litellm","chat_template_applied":true,"generation(...TRUNCATED)
{ "backend": "litellm", "model_name": "hosted_vllm/TAUR-dev/M-AT_ours_sft-sft", "n_responses": 4 }
[ false, false, false, false ]
["\\(23 \\times 7 + 21 + 23 = 322\\)","\\( (23 \\times 21) - 161 = 322 \\)","(23 * 21 - 161)","No va(...TRUNCATED)
"[{\"extraction_type\": \"final\", \"total_spans_found\": 5, \"non_overlapping_spans\": 5, \"span_st(...TRUNCATED)
"[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"\\\\(23 (...TRUNCATED)
[[false,false,false,false,false],[false,false,false],[false,false,false,false,false,false,false],[fa(...TRUNCATED)
[["\\( 23 * 21 + 7 - 68 = 322 \\)","None of the proposed methods resulted in exactly 322. It is math(...TRUNCATED)
"[[{\"pattern_used\": \"answer tag\", \"confidence\": 1.0, \"position\": 1569, \"span_text\": \"\\\\(...TRUNCATED)
"[[{\"is_correct\": false, \"error\": \"Could not evaluate expression\", \"answer_block\": \"\\\\( 2(...TRUNCATED)
{"flips_by":[0,0,0,0],"flips_total":0,"num_correct":0,"pass_at_n":0,"percent_correct":0.0,"skill_cou(...TRUNCATED)
2025-11-10
test
main
TAUR-dev/M-AT_ours_sft-sft
0
eval_sft
1
2025-11-10T02:50:48.472447
TAUR-dev/D-EVAL__standard_eval_v3__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg-eval_sft
End of preview. Expand in Data Studio

Experiment Tracker: FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg

Experiment Description: Evaluation experiment for task countdown_3arg from FinEval_16k_fulleval_AT_OURS-SFT

Start Time: 2025-11-10T01:57:50.540633

Tracker Dataset: TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1

Stages Completed

Total stages: 1

Models Created

Dataset Configurations

This tracker dataset contains the following configurations with immediate upload as stages complete:

Training Data (Complete Datasets)

Hyperparameters (Complete Configurations)

Logs (Stage-Specific)

Evaluation Results (Complete with Annotations)

Metadata

  • experiment_metadata: Timeline and stage information

Usage

Load specific configurations with:

from datasets import load_dataset

# Load experiment metadata
metadata = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'experiment_metadata')

# Load complete training datasets
sft_data = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'training_data__sft')
sft_metadata = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'training_data__sft_metadata')

# Load complete configurations
sft_hyperparams = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'hyperparameters__sft')
rl_hyperparams = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'hyperparameters__rl')

# Load stage-specific logs
sft_logs = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'logs__sft')
rl_logs = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'logs__rl')

# Load evaluation results with annotations
sft_eval_results = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'evals_eval_sft')
rl_eval_results = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg__v1', 'evals_eval_rl')

Models

Registry

All models from this experiment are automatically registered in the SkillFactory Model Registry with:

  • Complete training configuration (hyperparameters, datasets, methods)
  • Experiment lineage (links back to this tracker dataset)
  • Stage-specific metadata (SFT vs RL training details)
  • Structured input data references (training datasets and configurations)

Registry entries follow the naming pattern: Model - FinEval_16k_fulleval_AT_OURS-SFT-countdown_3arg - {stage_name} - {SFT/RL}


Generated by SkillFactory Experiment Management System All artifacts uploaded immediately as stages complete with perfect data provenance

Downloads last month
3