| [ |
| { |
| "name": "inspect_evals/apps - Sample 0", |
| "description": "Task from inspect_evals/apps dataset, sample 0", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nAn accordion is a string (yes, in the real world accordions are musical instruments, but let's forget about it for a while) which can be represented as a concatenation of: an opening bracket (ASCII code $091$), a colon (ASCII code $058$), some (possibly zero) vertical line characters (ASCII code $124$), another colon, and a closing bracket (ASCII code $093$). The length of the accordion is the number of characters in it.\n\nFor example, [::], [:||:] and [:|||:] are accordions having length $4$, $6$ and $7$. (:|:), {:||:}, [:], ]:||:[ are not accordions. \n\nYou are given a string $s$. You want to transform it into an accordion by removing some (possibly zero) characters from it. Note that you may not insert new characters or reorder existing ones. Is it possible to obtain an accordion by removing characters from $s$, and if so, what is the maximum possible length of the result?\n\n\n-----Input-----\n\nThe only line contains one string $s$ ($1 \\le |s| \\le 500000$). It consists of lowercase Latin letters and characters [, ], : and |.\n\n\n-----Output-----\n\nIf it is not possible to obtain an accordion by removing some characters from $s$, print $-1$. Otherwise print maximum possible length of the resulting accordion.\n\n\n-----Examples-----\nInput\n|[a:b:|]\n\nOutput\n4\n\nInput\n|]:[|:]\n\nOutput\n-1\n\n## Test Cases:\n```python\nassert solution('|[a:b:|]\\n') == '4\\n'\nassert solution('|]:[|:]\\n') == '-1\\n'\nassert solution(':][:\\n') == '-1\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": "0", |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 1", |
| "description": "Task from inspect_evals/apps dataset, sample 1", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nAnton has the integer x. He is interested what positive integer, which doesn't exceed x, has the maximum sum of digits.\n\nYour task is to help Anton and to find the integer that interests him. If there are several such integers, determine the biggest of them. \n\n\n-----Input-----\n\nThe first line contains the positive integer x (1 \u2264 x \u2264 10^18) \u2014 the integer which Anton has. \n\n\n-----Output-----\n\nPrint the positive integer which doesn't exceed x and has the maximum sum of digits. If there are several such integers, print the biggest of them. Printed integer must not contain leading zeros.\n\n\n-----Examples-----\nInput\n100\n\nOutput\n99\n\nInput\n48\n\nOutput\n48\n\nInput\n521\n\nOutput\n499\n\n## Test Cases:\n```python\nassert solution('100\\n') == '99\\n'\nassert solution('48\\n') == '48\\n'\nassert solution('521\\n') == '499\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 1, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 2", |
| "description": "Task from inspect_evals/apps dataset, sample 2", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nApart from having lots of holidays throughout the year, residents of Berland also have whole lucky years. Year is considered lucky if it has no more than 1 non-zero digit in its number. So years 100, 40000, 5 are lucky and 12, 3001 and 12345 are not.\n\nYou are given current year in Berland. Your task is to find how long will residents of Berland wait till the next lucky year.\n\n\n-----Input-----\n\nThe first line contains integer number n (1 \u2264 n \u2264 10^9) \u2014 current year in Berland.\n\n\n-----Output-----\n\nOutput amount of years from the current year to the next lucky one.\n\n\n-----Examples-----\nInput\n4\n\nOutput\n1\n\nInput\n201\n\nOutput\n99\n\nInput\n4000\n\nOutput\n1000\n\n\n\n-----Note-----\n\nIn the first example next lucky year is 5. In the second one \u2014 300. In the third \u2014 5000.\n\n## Test Cases:\n```python\nassert solution('4\\n') == '1\\n'\nassert solution('201\\n') == '99\\n'\nassert solution('4000\\n') == '1000\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 2, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 3", |
| "description": "Task from inspect_evals/apps dataset, sample 3", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nYou have a long fence which consists of $n$ sections. Unfortunately, it is not painted, so you decided to hire $q$ painters to paint it. $i$-th painter will paint all sections $x$ such that $l_i \\le x \\le r_i$.\n\nUnfortunately, you are on a tight budget, so you may hire only $q - 2$ painters. Obviously, only painters you hire will do their work.\n\nYou want to maximize the number of painted sections if you choose $q - 2$ painters optimally. A section is considered painted if at least one painter paints it.\n\n\n-----Input-----\n\nThe first line contains two integers $n$ and $q$ ($3 \\le n, q \\le 5000$) \u2014 the number of sections and the number of painters availible for hire, respectively.\n\nThen $q$ lines follow, each describing one of the painters: $i$-th line contains two integers $l_i$ and $r_i$ ($1 \\le l_i \\le r_i \\le n$).\n\n\n-----Output-----\n\nPrint one integer \u2014 maximum number of painted sections if you hire $q - 2$ painters.\n\n\n-----Examples-----\nInput\n7 5\n1 4\n4 5\n5 6\n6 7\n3 5\n\nOutput\n7\n\nInput\n4 3\n1 1\n2 2\n3 4\n\nOutput\n2\n\nInput\n4 4\n1 1\n2 2\n2 3\n3 4\n\nOutput\n3\n\n## Test Cases:\n```python\nassert solution('7 5\\n1 4\\n4 5\\n5 6\\n6 7\\n3 5\\n') == '7\\n'\nassert solution('4 3\\n1 1\\n2 2\\n3 4\\n') == '2\\n'\nassert solution('4 4\\n1 1\\n2 2\\n2 3\\n3 4\\n') == '3\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 3, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 4", |
| "description": "Task from inspect_evals/apps dataset, sample 4", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nJamie loves sleeping. One day, he decides that he needs to wake up at exactly hh: mm. However, he hates waking up, so he wants to make waking up less painful by setting the alarm at a lucky time. He will then press the snooze button every x minutes until hh: mm is reached, and only then he will wake up. He wants to know what is the smallest number of times he needs to press the snooze button.\n\nA time is considered lucky if it contains a digit '7'. For example, 13: 07 and 17: 27 are lucky, while 00: 48 and 21: 34 are not lucky.\n\nNote that it is not necessary that the time set for the alarm and the wake-up time are on the same day. It is guaranteed that there is a lucky time Jamie can set so that he can wake at hh: mm.\n\nFormally, find the smallest possible non-negative integer y such that the time representation of the time x\u00b7y minutes before hh: mm contains the digit '7'.\n\nJamie uses 24-hours clock, so after 23: 59 comes 00: 00.\n\n\n-----Input-----\n\nThe first line contains a single integer x (1 \u2264 x \u2264 60).\n\nThe second line contains two two-digit integers, hh and mm (00 \u2264 hh \u2264 23, 00 \u2264 mm \u2264 59).\n\n\n-----Output-----\n\nPrint the minimum number of times he needs to press the button.\n\n\n-----Examples-----\nInput\n3\n11 23\n\nOutput\n2\n\nInput\n5\n01 07\n\nOutput\n0\n\n\n\n-----Note-----\n\nIn the first sample, Jamie needs to wake up at 11:23. So, he can set his alarm at 11:17. He would press the snooze button when the alarm rings at 11:17 and at 11:20.\n\nIn the second sample, Jamie can set his alarm at exactly at 01:07 which is lucky.\n\n## Test Cases:\n```python\nassert solution('3\\n11 23\\n') == '2\\n'\nassert solution('5\\n01 07\\n') == '0\\n'\nassert solution('34\\n09 24\\n') == '3\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 4, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 5", |
| "description": "Task from inspect_evals/apps dataset, sample 5", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nLuba is surfing the Internet. She currently has n opened tabs in her browser, indexed from 1 to n from left to right. The mouse cursor is currently located at the pos-th tab. Luba needs to use the tabs with indices from l to r (inclusive) for her studies, and she wants to close all the tabs that don't belong to this segment as fast as possible.\n\nEach second Luba can either try moving the cursor to the left or to the right (if the cursor is currently at the tab i, then she can move it to the tab max(i - 1, a) or to the tab min(i + 1, b)) or try closing all the tabs to the left or to the right of the cursor (if the cursor is currently at the tab i, she can close all the tabs with indices from segment [a, i - 1] or from segment [i + 1, b]). In the aforementioned expressions a and b denote the minimum and maximum index of an unclosed tab, respectively. For example, if there were 7 tabs initially and tabs 1, 2 and 7 are closed, then a = 3, b = 6.\n\nWhat is the minimum number of seconds Luba has to spend in order to leave only the tabs with initial indices from l to r inclusive opened?\n\n\n-----Input-----\n\nThe only line of input contains four integer numbers n, pos, l, r (1 \u2264 n \u2264 100, 1 \u2264 pos \u2264 n, 1 \u2264 l \u2264 r \u2264 n) \u2014 the number of the tabs, the cursor position and the segment which Luba needs to leave opened.\n\n\n-----Output-----\n\nPrint one integer equal to the minimum number of seconds required to close all the tabs outside the segment [l, r].\n\n\n-----Examples-----\nInput\n6 3 2 4\n\nOutput\n5\n\nInput\n6 3 1 3\n\nOutput\n1\n\nInput\n5 2 1 5\n\nOutput\n0\n\n\n\n-----Note-----\n\nIn the first test Luba can do the following operations: shift the mouse cursor to the tab 2, close all the tabs to the left of it, shift the mouse cursor to the tab 3, then to the tab 4, and then close all the tabs to the right of it.\n\nIn the second test she only needs to close all the tabs to the right of the current position of the cursor.\n\nIn the third test Luba doesn't need to do anything.\n\n## Test Cases:\n```python\nassert solution('6 3 2 4\\n') == '5\\n'\nassert solution('6 3 1 3\\n') == '1\\n'\nassert solution('5 2 1 5\\n') == '0\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 5, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 6", |
| "description": "Task from inspect_evals/apps dataset, sample 6", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nYou are fighting with Zmei Gorynich \u2014 a ferocious monster from Slavic myths, a huge dragon-like reptile with multiple heads! \n\n $m$ \n\nInitially Zmei Gorynich has $x$ heads. You can deal $n$ types of blows. If you deal a blow of the $i$-th type, you decrease the number of Gorynich's heads by $min(d_i, curX)$, there $curX$ is the current number of heads. But if after this blow Zmei Gorynich has at least one head, he grows $h_i$ new heads. If $curX = 0$ then Gorynich is defeated. \n\nYou can deal each blow any number of times, in any order.\n\nFor example, if $curX = 10$, $d = 7$, $h = 10$ then the number of heads changes to $13$ (you cut $7$ heads off, but then Zmei grows $10$ new ones), but if $curX = 10$, $d = 11$, $h = 100$ then number of heads changes to $0$ and Zmei Gorynich is considered defeated.\n\nCalculate the minimum number of blows to defeat Zmei Gorynich!\n\nYou have to answer $t$ independent queries.\n\n\n-----Input-----\n\nThe first line contains one integer $t$ ($1 \\le t \\le 100$) \u2013 the number of queries.\n\nThe first line of each query contains two integers $n$ and $x$ ($1 \\le n \\le 100$, $1 \\le x \\le 10^9$) \u2014 the number of possible types of blows and the number of heads Zmei initially has, respectively.\n\nThe following $n$ lines of each query contain the descriptions of types of blows you can deal. The $i$-th line contains two integers $d_i$ and $h_i$ ($1 \\le d_i, h_i \\le 10^9$) \u2014 the description of the $i$-th blow.\n\n\n-----Output-----\n\nFor each query print the minimum number of blows you have to deal to defeat Zmei Gorynich. \n\nIf Zmei Gorynuch cannot be defeated print $-1$.\n\n\n-----Example-----\nInput\n3\n3 10\n6 3\n8 2\n1 4\n4 10\n4 1\n3 2\n2 6\n1 100\n2 15\n10 11\n14 100\n\nOutput\n2\n3\n-1\n\n\n\n-----Note-----\n\nIn the first query you can deal the first blow (after that the number of heads changes to $10 - 6 + 3 = 7$), and then deal the second blow.\n\nIn the second query you just deal the first blow three times, and Zmei is defeated. \n\nIn third query you can not defeat Zmei Gorynich. Maybe it's better to convince it to stop fighting?\n\n## Test Cases:\n```python\nassert solution('3\\n3 10\\n6 3\\n8 2\\n1 4\\n4 10\\n4 1\\n3 2\\n2 6\\n1 100\\n2 15\\n10 11\\n14 100\\n') == '2\\n3\\n-1\\n'\nassert solution('7\\n5 1000000000\\n2 1\\n1 10\\n1 1\\n4 1000000000\\n3 3\\n1 1000000000\\n5 1\\n2 999999999\\n3 1\\n2 10000000\\n4 10000000\\n10000000 999999999\\n9999900 12\\n9999999 55\\n9999999 1\\n2 1000000\\n1000000 1000000\\n999999 1\\n3 999999911\\n3 1\\n11 1000000000\\n10 9\\n3 1000000000\\n1231 1200\\n1000 800\\n1 100\\n') == '999999997\\n250000000\\n499999999\\n1\\n1\\n499999951\\n4999995\\n'\nassert solution('1\\n1 1\\n3 1\\n') == '1\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 6, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 7", |
| "description": "Task from inspect_evals/apps dataset, sample 7", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nAnton likes to listen to fairy tales, especially when Danik, Anton's best friend, tells them. Right now Danik tells Anton a fairy tale:\n\n\"Once upon a time, there lived an emperor. He was very rich and had much grain. One day he ordered to build a huge barn to put there all his grain. Best builders were building that barn for three days and three nights. But they overlooked and there remained a little hole in the barn, from which every day sparrows came through. Here flew a sparrow, took a grain and flew away...\"\n\nMore formally, the following takes place in the fairy tale. At the beginning of the first day the barn with the capacity of n grains was full. Then, every day (starting with the first day) the following happens: m grains are brought to the barn. If m grains doesn't fit to the barn, the barn becomes full and the grains that doesn't fit are brought back (in this problem we can assume that the grains that doesn't fit to the barn are not taken into account). Sparrows come and eat grain. In the i-th day i sparrows come, that is on the first day one sparrow come, on the second day two sparrows come and so on. Every sparrow eats one grain. If the barn is empty, a sparrow eats nothing. \n\nAnton is tired of listening how Danik describes every sparrow that eats grain from the barn. Anton doesn't know when the fairy tale ends, so he asked you to determine, by the end of which day the barn will become empty for the first time. Help Anton and write a program that will determine the number of that day!\n\n\n-----Input-----\n\nThe only line of the input contains two integers n and m (1 \u2264 n, m \u2264 10^18)\u00a0\u2014 the capacity of the barn and the number of grains that are brought every day.\n\n\n-----Output-----\n\nOutput one integer\u00a0\u2014 the number of the day when the barn will become empty for the first time. Days are numbered starting with one.\n\n\n-----Examples-----\nInput\n5 2\n\nOutput\n4\n\nInput\n8 1\n\nOutput\n5\n\n\n\n-----Note-----\n\nIn the first sample the capacity of the barn is five grains and two grains are brought every day. The following happens: At the beginning of the first day grain is brought to the barn. It's full, so nothing happens. At the end of the first day one sparrow comes and eats one grain, so 5 - 1 = 4 grains remain. At the beginning of the second day two grains are brought. The barn becomes full and one grain doesn't fit to it. At the end of the second day two sparrows come. 5 - 2 = 3 grains remain. At the beginning of the third day two grains are brought. The barn becomes full again. At the end of the third day three sparrows come and eat grain. 5 - 3 = 2 grains remain. At the beginning of the fourth day grain is brought again. 2 + 2 = 4 grains remain. At the end of the fourth day four sparrows come and eat grain. 4 - 4 = 0 grains remain. The barn is empty. \n\nSo the answer is 4, because by the end of the fourth day the barn becomes empty.\n\n## Test Cases:\n```python\nassert solution('5 2\\n') == '4\\n'\nassert solution('8 1\\n') == '5\\n'\nassert solution('32 5\\n') == '12\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 7, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 8", |
| "description": "Task from inspect_evals/apps dataset, sample 8", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nTokitsukaze is playing a game derivated from Japanese mahjong. In this game, she has three tiles in her hand. Each tile she owns is a suited tile, which means it has a suit (manzu, pinzu or souzu) and a number (a digit ranged from $1$ to $9$). In this problem, we use one digit and one lowercase letter, which is the first character of the suit, to represent a suited tile. All possible suited tiles are represented as 1m, 2m, $\\ldots$, 9m, 1p, 2p, $\\ldots$, 9p, 1s, 2s, $\\ldots$, 9s.\n\nIn order to win the game, she must have at least one mentsu (described below) in her hand, so sometimes she should draw extra suited tiles. After drawing a tile, the number of her tiles increases by one. She can draw any tiles she wants, including those already in her hand.\n\nDo you know the minimum number of extra suited tiles she needs to draw so that she can win?\n\nHere are some useful definitions in this game: A mentsu, also known as meld, is formed by a koutsu or a shuntsu; A koutsu, also known as triplet, is made of three identical tiles, such as [1m, 1m, 1m], however, [1m, 1p, 1s] or [1m, 4m, 7m] is NOT a koutsu; A shuntsu, also known as sequence, is made of three sequential numbered tiles in the same suit, such as [1m, 2m, 3m] and [5s, 7s, 6s], however, [9m, 1m, 2m] or [1m, 2p, 3s] is NOT a shuntsu. \n\nSome examples: [2m, 3p, 2s, 4m, 1s, 2s, 4s] \u2014 it contains no koutsu or shuntsu, so it includes no mentsu; [4s, 3m, 3p, 4s, 5p, 4s, 5p] \u2014 it contains a koutsu, [4s, 4s, 4s], but no shuntsu, so it includes a mentsu; [5p, 5s, 9m, 4p, 1s, 7p, 7m, 6p] \u2014 it contains no koutsu but a shuntsu, [5p, 4p, 6p] or [5p, 7p, 6p], so it includes a mentsu. \n\nNote that the order of tiles is unnecessary and you can assume the number of each type of suited tiles she can draw is infinite.\n\n\n-----Input-----\n\nThe only line contains three strings\u00a0\u2014 the tiles in Tokitsukaze's hand. For each string, the first character is a digit ranged from $1$ to $9$ and the second character is m, p or s.\n\n\n-----Output-----\n\nPrint a single integer\u00a0\u2014 the minimum number of extra suited tiles she needs to draw.\n\n\n-----Examples-----\nInput\n1s 2s 3s\n\nOutput\n0\n\nInput\n9m 9m 9m\n\nOutput\n0\n\nInput\n3p 9m 2p\n\nOutput\n1\n\n\n\n-----Note-----\n\nIn the first example, Tokitsukaze already has a shuntsu.\n\nIn the second example, Tokitsukaze already has a koutsu.\n\nIn the third example, Tokitsukaze can get a shuntsu by drawing one suited tile\u00a0\u2014 1p or 4p. The resulting tiles will be [3p, 9m, 2p, 1p] or [3p, 9m, 2p, 4p].\n\n## Test Cases:\n```python\nassert solution('1s 2s 3s\\n') == '0\\n'\nassert solution('9m 9m 9m\\n') == '0\\n'\nassert solution('3p 9m 2p\\n') == '1\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 8, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| }, |
| { |
| "name": "inspect_evals/apps - Sample 9", |
| "description": "Task from inspect_evals/apps dataset, sample 9", |
| "prompt": "\n# Now, complete the following task.\n\n## Question:\nYet another round on DecoForces is coming! Grandpa Maks wanted to participate in it but someone has stolen his precious sofa! And how can one perform well with such a major loss?\n\nFortunately, the thief had left a note for Grandpa Maks. This note got Maks to the sofa storehouse. Still he had no idea which sofa belongs to him as they all looked the same!\n\nThe storehouse is represented as matrix n \u00d7 m. Every sofa takes two neighbouring by some side cells. No cell is covered by more than one sofa. There can be empty cells.\n\nSofa A is standing to the left of sofa B if there exist two such cells a and b that x_{a} < x_{b}, a is covered by A and b is covered by B. Sofa A is standing to the top of sofa B if there exist two such cells a and b that y_{a} < y_{b}, a is covered by A and b is covered by B. Right and bottom conditions are declared the same way. \n\nNote that in all conditions A \u2260 B. Also some sofa A can be both to the top of another sofa B and to the bottom of it. The same is for left and right conditions.\n\nThe note also stated that there are cnt_{l} sofas to the left of Grandpa Maks's sofa, cnt_{r} \u2014 to the right, cnt_{t} \u2014 to the top and cnt_{b} \u2014 to the bottom.\n\nGrandpa Maks asks you to help him to identify his sofa. It is guaranteed that there is no more than one sofa of given conditions.\n\nOutput the number of Grandpa Maks's sofa. If there is no such sofa that all the conditions are met for it then output -1.\n\n\n-----Input-----\n\nThe first line contains one integer number d (1 \u2264 d \u2264 10^5) \u2014 the number of sofas in the storehouse.\n\nThe second line contains two integer numbers n, m (1 \u2264 n, m \u2264 10^5) \u2014 the size of the storehouse.\n\nNext d lines contains four integer numbers x_1, y_1, x_2, y_2 (1 \u2264 x_1, x_2 \u2264 n, 1 \u2264 y_1, y_2 \u2264 m) \u2014 coordinates of the i-th sofa. It is guaranteed that cells (x_1, y_1) and (x_2, y_2) have common side, (x_1, y_1) \u2260 (x_2, y_2) and no cell is covered by more than one sofa.\n\nThe last line contains four integer numbers cnt_{l}, cnt_{r}, cnt_{t}, cnt_{b} (0 \u2264 cnt_{l}, cnt_{r}, cnt_{t}, cnt_{b} \u2264 d - 1).\n\n\n-----Output-----\n\nPrint the number of the sofa for which all the conditions are met. Sofas are numbered 1 through d as given in input. If there is no such sofa then print -1.\n\n\n-----Examples-----\nInput\n2\n3 2\n3 1 3 2\n1 2 2 2\n1 0 0 1\n\nOutput\n1\n\nInput\n3\n10 10\n1 2 1 1\n5 5 6 5\n6 4 5 4\n2 1 2 0\n\nOutput\n2\n\nInput\n2\n2 2\n2 1 1 1\n1 2 2 2\n1 0 0 0\n\nOutput\n-1\n\n\n\n-----Note-----\n\nLet's consider the second example. The first sofa has 0 to its left, 2 sofas to its right ((1, 1) is to the left of both (5, 5) and (5, 4)), 0 to its top and 2 to its bottom (both 2nd and 3rd sofas are below). The second sofa has cnt_{l} = 2, cnt_{r} = 1, cnt_{t} = 2 and cnt_{b} = 0. The third sofa has cnt_{l} = 2, cnt_{r} = 1, cnt_{t} = 1 and cnt_{b} = 1. \n\nSo the second one corresponds to the given conditions.\n\nIn the third example The first sofa has cnt_{l} = 1, cnt_{r} = 1, cnt_{t} = 0 and cnt_{b} = 1. The second sofa has cnt_{l} = 1, cnt_{r} = 1, cnt_{t} = 1 and cnt_{b} = 0. \n\nAnd there is no sofa with the set (1, 0, 0, 0) so the answer is -1.\n\n## Test Cases:\n```python\nassert solution('2\\n3 2\\n3 1 3 2\\n1 2 2 2\\n1 0 0 1\\n') == '1\\n'\nassert solution('3\\n10 10\\n1 2 1 1\\n5 5 6 5\\n6 4 5 4\\n2 1 2 0\\n') == '2\\n'\nassert solution('2\\n2 2\\n2 1 1 1\\n1 2 2 2\\n1 0 0 0\\n') == '-1\\n'\n```\n\n## Completion:\n", |
| "mcp_config": { |
| "inspect": { |
| "command": "docker", |
| "args": [ |
| "run", |
| "--rm", |
| "-i", |
| "--init", |
| "-v", |
| "/var/run/docker.sock:/var/run/docker.sock", |
| "-v", |
| "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals", |
| "hud-inspect:latest" |
| ] |
| } |
| }, |
| "setup_tool": { |
| "name": "load_task", |
| "arguments": { |
| "task_name": "inspect_evals/apps", |
| "sample_id": 9, |
| "model": "claude-sonnet-4-20250514" |
| } |
| }, |
| "evaluate_tool": { |
| "name": "score_task", |
| "arguments": {} |
| }, |
| "agent_config": { |
| "agent_class": "server.agents.InspectAgent", |
| "disallowed_tools": [ |
| "_store_agent_message", |
| "load_task", |
| "score_task" |
| ], |
| "allowed_tools": [], |
| "initial_screenshot": false |
| }, |
| "system_prompt": "You are an expert Python programmer. You will be given a task and a test cases. Be mindful of the test case format.\nIf the test case expects a string and you return an integer, you will fail the test case. There are additional test cases not provided that will be used to verify your solution. These are very specific, and your solution should account for edge cases. Write the complete Python function to solve the task. Do not add whitespace before or after the code block. Only use imports that are included in Python's standard library.\n\nBefore answering, reason in a step-by-step manner as to get the right answer. Then write the code block in the following format:\n\n```python\n[code block]\n```\n\nAfter writing the code block, please do not add any additional text or rationalization.\n\nFor instance:\n ## Question:\n Write a function that takes a string of space-separated integers and returns the sum of all numbers that are greater than 5.\nExample:\nInput: \"1 6 2 7 3 8\"\nOutput: 21 (because 6 + 7 + 8 = 21)\n\n ## Test Cases:\n ```python\nassert solution(\"1 6 2 7 3 8\") == 21\nassert solution(\"1 2 3 4 5\") == 0\nassert solution(\"10 20 30\") == 60\n ```\n\n ## Completion:\n Let me solve this step by step:\n\n1. Split the input string into a list of integers\n2. Filter numbers greater than 5\n3. Sum the filtered numbers\n\n```python\ndef solution(s):\n return sum(int(x) for x in s.split() if int(x) > 5)\n```" |
| } |
| ] |
