Dataset Viewer
Auto-converted to Parquet Duplicate
problem
stringlengths
20
1.11k
source
stringclasses
4 values
requested_effort
stringclasses
1 value
thinking
stringlengths
683
106k
thinking_tokens
int64
280
32k
response
stringlengths
117
3.37k
predicted_answer
stringclasses
313 values
ground_truth
stringclasses
291 values
correct
bool
1 class
finish_reason
stringclasses
1 value
effort
stringclasses
3 values
The digits 2, 4, 6, and 8 are each used once to create two 2-digit numbers. What is the largest possible difference between the two 2-digit numbers? (A) 66 (B) 62 (C) 58 (D) 44 (E) 36
olympiads
medium
The user wants to find the largest possible difference between two 2-digit numbers created using the digits 2, 4, 6, and 8, each exactly once. Let the two 2-digit numbers be $A$ and $B$. The digits available are $\{2, 4, 6, 8\}$. We need to form two numbers, say $d_1 d_2$ and $d_3 d_4$, where $d_1, d_2, d_3, d_4$ is a...
6,278
To find the largest possible difference between two 2-digit numbers created using the digits 2, 4, 6, and 8 (each used exactly once), we should follow a systematic approach: 1. **Identify the structure:** Let the two 2-digit numbers be $A$ and $B$. We want to maximize the difference $A - B$ (assuming $A > B$). ...
62
62
true
stop
low
The digits 2, 4, 6, and 8 are each used once to create two 2-digit numbers. What is the largest possible difference between the two 2-digit numbers? (A) 66 (B) 62 (C) 58 (D) 44 (E) 36
olympiads
medium
Here's a thinking process that leads to the solution: 1. **Understand the Goal:** The objective is to find the *largest possible difference* between two 2-digit numbers formed using the digits $\{2, 4, 6, 8\}$ exactly once each. 2. **Analyze the Components:** * Available digits: 2, 4, 6, 8. * Number of ...
7,022
To find the largest possible difference between two 2-digit numbers formed using the digits 2, 4, 6, and 8 exactly once, we need to maximize the value of the first number and minimize the value of the second number. Let the two numbers be $N_1$ and $N_2$. We want to maximize $N_1 - N_2$. A 2-digit number can be repr...
62
62
true
stop
medium
The digits 2, 4, 6, and 8 are each used once to create two 2-digit numbers. What is the largest possible difference between the two 2-digit numbers? (A) 66 (B) 62 (C) 58 (D) 44 (E) 36
olympiads
medium
The problem asks for the largest possible difference between two 2-digit numbers formed using the digits 2, 4, 6, and 8, where each digit is used exactly once. Let the two 2-digit numbers be $A$ and $B$. We want to maximize $|A - B|$. Let's assume $A > B$, so we want to maximize $A - B$. The digits available are $\{2,...
18,386
To find the largest possible difference between two 2-digit numbers created using the digits 2, 4, 6, and 8 exactly once, we let the numbers be $A = 10a_1 + a_2$ and $B = 10b_1 + b_2$. The difference is: $$A - B = 10(a_1 - b_1) + (a_2 - b_2)$$ To maximize this difference, we should prioritize the tens place because ...
62
62
true
stop
high
On his birthday, the last guest to arrive was Yana, who gave Andrey a ball, and the second last was Eduard, who gave him a calculator. While using the calculator, Andrey noticed that the product of the total number of gifts he received and the number of gifts he had before Eduard arrived is exactly 16 more than the pro...
olympiads
medium
Here's a thinking process that leads to the solution: 1. **Analyze the Problem Statement:** * **Characters:** Andrey (the birthday boy), Yana (last guest), Eduard (second last guest). * **Events:** * Yana gave a ball. * Eduard gave a calculator. * Andrey had some number of g...
5,238
To find the total number of gifts Andrey has, we can set up an algebraic equation based on the information provided: 1. **Define the Variables:** * Let $n$ be the total number of gifts Andrey received. * Let $a$ be Andrey's age. * Because Eduard was the **second-to-last** guest, Andrey had receive...
18
18
true
stop
low
On his birthday, the last guest to arrive was Yana, who gave Andrey a ball, and the second last was Eduard, who gave him a calculator. While using the calculator, Andrey noticed that the product of the total number of gifts he received and the number of gifts he had before Eduard arrived is exactly 16 more than the pro...
olympiads
medium
Here's a thinking process that leads to the solution: 1. **Analyze the Problem Statement:** * **Characters:** Andrey (the birthday boy), Yana, Eduard. * **Events:** * Eduard arrived 2nd to last. * Yana arrived last. * **Gifts:** * Eduard gave a calculator (1 gift). ...
15,784
Let $N$ be the total number of gifts Andrey received. Let $A$ be Andrey's age. Let $x$ be the number of gifts Andrey had before Eduard arrived. Based on the problem statement, we can determine the number of gifts at different stages: 1. **Before Eduard arrived:** Andrey had $x$ gifts. 2. **Eduard's arrival:** Edua...
18
18
true
stop
medium
"On his birthday, the last guest to arrive was Yana, who gave Andrey a ball, and the second last was(...TRUNCATED)
olympiads
medium
"Let $n$ be the number of gifts Andrey has *before* Eduard arrives.\nThe sequence of events is descr(...TRUNCATED)
16,790
"\n\nLet $n$ be the number of gifts Andrey had before Eduard arrived.\nAccording to the problem:\n- (...TRUNCATED)
18
18
true
stop
high
"Thomas has constant speeds for both running and walking. When a down-escalator is moving, Thomas ca(...TRUNCATED)
olympiads
medium
"Here's a thinking process that leads to the solution:\n\n1. **Understand the Goal:** The problem a(...TRUNCATED)
5,491
"\n\nTo find the time it takes for Thomas to walk down the stationary escalator, we can analyze the (...TRUNCATED)
60
60
true
stop
low
"Thomas has constant speeds for both running and walking. When a down-escalator is moving, Thomas ca(...TRUNCATED)
olympiads
medium
"Here's a thinking process to solve the problem:\n\n1. **Understand the Goal:** The objective is to(...TRUNCATED)
6,228
"\n\nLet $D$ be the length of the escalator.\nLet $r$ be Thomas's running speed relative to the esca(...TRUNCATED)
60
60
true
stop
medium
"Thomas has constant speeds for both running and walking. When a down-escalator is moving, Thomas ca(...TRUNCATED)
olympiads
medium
"Here's a thinking process to solve the problem:\n\n1. **Understand the Goal:** The objective is to(...TRUNCATED)
7,082
"\n\nLet $D$ be the length of the escalator.\nLet $v_r$ be Thomas's running speed.\nLet $v_w$ be Tho(...TRUNCATED)
60
60
true
stop
high
"Eighty percent of dissatisfied customers leave angry reviews about a certain online store. Among sa(...TRUNCATED)
olympiads
medium
"Here's a thinking process that leads to the solution:\n\n1. **Understand the Goal:** The objective(...TRUNCATED)
6,460
"\n\nTo estimate the probability that the next customer will be satisfied, we can work backward from(...TRUNCATED)
0.64
0.64
true
stop
low
End of preview. Expand in Data Studio

Sherry Reasoning Effort Dataset (0.1)

Math reasoning traces at 3 effort levels (low, medium, high) generated by Qwen3.6-35B-A3B (NVFP4) via API.

Description

Each problem was sampled from NuminaMath-CoT (numeric-answer problems only, proofs filtered out). For every problem, 3 candidates were sampled from the generator model with thinking enabled, verified against the ground truth, and the 3 verified traces with different reasoning depths were kept: the shortest labeled low, the median medium, and the longest high.

Source

  • Problems: AI-MO/NuminaMath-CoT (Apache-2.0)
  • Generator: unsloth/Qwen3.6-35B-A3B-NVFP4-Fast (OpenAI-compatible API)
  • Verification: exact-match/numeric on \boxed{} against NuminaMath ground truth (all traces correct)

Format

Each row is a JSON object:

  • problem: the math problem text
  • source: NuminaMath source (olympiads, synthetic_math, cn_k12, orca_math)
  • effort: one of "low", "medium", "high" (assigned from observed reasoning depth)
  • thinking: the reasoning content (separated from response)
  • thinking_tokens: actual reasoning tokens reported by the API
  • response: the final response (contains \boxed{})
  • predicted_answer: extracted answer from response
  • ground_truth: extracted answer from NuminaMath solution
  • correct: whether predicted matches ground truth (always true)
  • finish_reason: API finish reason (always "stop"; truncated candidates discarded)
  • requested_effort: prompt used for candidate generation (uniform "medium")

Intended Use

SFT training for reasoning effort control in language models.

Downloads last month
22