Dataset Viewer
Auto-converted to Parquet Duplicate
uuid
large_stringclasses
1 value
expected_answer
large_stringclasses
2 values
problem
large_stringclasses
2 values
original_expected_answer
large_stringclasses
2 values
changed_answer_to_majority
bool
1 class
data_source
large_stringclasses
1 value
messages
listlengths
4
4
used_in
listlengths
1
1
metadata
dict
license
large_stringclasses
1 value
tools
listlengths
1
1
url
float64
user_name
float64
user_url
float64
\( (x, y) = (2^{998}, 2^{998}) \)
Find a pair of integers $(x, y)$ such that $15x^2 + y^2 = 2^{2000}$.
\( (x, y) = (2^{998}, 2^{998}) \)
false
aops
[ { "content": "Solve the following math problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nFind a pair of integers $(x, y)$ such that $15x^2 + y^2 = 2^{2000}$.", "name": "", "reasoning_content": "", "role": "user", "tool_call_id": "", "tool_calls": [] }, { "co...
[ "nano_v3" ]
{ "reason_high_no_tool": { "accuracy": 0.125, "count": 8, "pass": 1 }, "reason_high_with_tool": { "accuracy": 0, "count": 8, "pass": 0 }, "reason_low_no_tool": { "accuracy": 0.25, "count": 8, "pass": 2 }, "reason_low_with_tool": { "accuracy": 0.25, "count": 8, ...
cc-by-4.0
[ { "function": { "description": "Call this function to execute Python code in a stateful Jupyter notebook environment. Python will respond with the output of the execution or time out after 120.0 seconds.", "name": "stateful_python_code_exec", "parameters": { "properties": { "...
null
null
null
\((a, a^2 - 1)\) for any integer \(a > 1\)
Find infinitely many pairs of integers \(a\) and \(b\) with \(1 < a < b\) such that \(ab\) exactly divides \(a^2 + b^2 - 1\).
\((a, a^2 - 1)\) for any integer \(a > 1\)
false
aops
[ { "content": "Solve the following math problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nFind infinitely many pairs of integers \\(a\\) and \\(b\\) with \\(1 < a < b\\) such that \\(ab\\) exactly divides \\(a^2 + b^2 - 1\\).", "name": "", "reasoning_content": "", "role": "u...
[ "nano_v3" ]
{ "reason_high_no_tool": { "accuracy": 0, "count": 8, "pass": 0 }, "reason_high_with_tool": { "accuracy": 0, "count": 8, "pass": 0 }, "reason_low_no_tool": { "accuracy": 0, "count": 8, "pass": 0 }, "reason_low_with_tool": { "accuracy": 0, "count": 8, "pass":...
cc-by-4.0
[ { "function": { "description": "Call this function to execute Python code in a stateful Jupyter notebook environment. Python will respond with the output of the execution or time out after 120.0 seconds.", "name": "stateful_python_code_exec", "parameters": { "properties": { "...
null
null
null

Nemotron-Math-v2 Truly Hard Tool Subset

Problems where ALL 6 reasoning regimes score <= 3/8 (37.5%). These are the hardest problems in Nemotron-Math-v2.

Splits

Split Accuracy Before (dedup) After (truly hard)
pass0of8 0/8 2 2
pass1of8 1/8 1,195 689
pass2of8 2/8 652 168
pass3of8 3/8 875 107
Total 2,724 966

Filter

All 6 regimes must have accuracy <= 0.375:

  • reason_high_with_tool / reason_high_no_tool
  • reason_medium_with_tool / reason_medium_no_tool
  • reason_low_with_tool / reason_low_no_tool
Downloads last month
17