[ { "name": "inspect_evals/mathvista - Sample 749", "description": "Task from inspect_evals/mathvista dataset, sample 749", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nAs shown in the figure, in \u25b3ABC, DE \u2225 BC, if AB = 7.0, AC = 5.0, AD = 3.0, then DE = ()\n\nA) \\frac{15}{4}cm\nB) \\frac{20}{3}cm\nC) \\frac{15}{7}cm\nD) \\frac{20}{7}cm", "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/mathvista", "sample_id": "749", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 270", "description": "Task from inspect_evals/mathvista dataset, sample 270", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nAs shown in the figure, points A and B are three points on \u2299O and AB = AC. Connect BO and CO, if \u2220ABC = 65.0, then the degree of \u2220BOC is ()\n\nA) 50\u00b0\nB) 65\u00b0\nC) 100\u00b0\nD) 130\u00b0", "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/mathvista", "sample_id": "270", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 912", "description": "Task from inspect_evals/mathvista dataset, sample 912", "prompt": "\nAnswer the following question. The entire content of your response should be of the following format: 'ANSWER: $ANSWER' (without quotes) where $ANSWER is your answer.\n\nHint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end.\nQuestion: How many people in the image were born after the end of World War II?\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/mathvista", "sample_id": "912", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 531", "description": "Task from inspect_evals/mathvista dataset, sample 531", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D,E.\n\nWhat time is shown? Answer by typing a time word, not a number. It is (_) to five.\n\nA) half\nB) quarter\nC) o'clock\nD) quarter to\nE) quarter past", "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/mathvista", "sample_id": "531", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 829", "description": "Task from inspect_evals/mathvista dataset, sample 829", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D,E.\n\nWhich matchstick needs to be moved in order to create a square?\n\nA) Top\nB) Bottom\nC) Left\nD) Right\nE) Not possible", "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/mathvista", "sample_id": "829", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 523", "description": "Task from inspect_evals/mathvista dataset, sample 523", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nAs shown in the figure, a teaching interest group wants to measure the height of a tree CD. They firstly measured the elevation angle of the tree top C at point A as 30.0, and then proceeded 10.0 along the direction of AD to point B, and the elevation angle of tree top C measured at B is 60.0 (the three points A, B, and D are on the same straight line), then the height of the tree CD is ()\n\nA) 10m\nB) 5m\nC) 5\u221a{3}m\nD) 10\u221a{3}m", "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/mathvista", "sample_id": "523", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 282", "description": "Task from inspect_evals/mathvista dataset, sample 282", "prompt": "\nAnswer the following question. The entire content of your response should be of the following format: 'ANSWER: $ANSWER' (without quotes) where $ANSWER is your answer.\n\nHint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end.\nQuestion: What percent of the sun is showing?\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/mathvista", "sample_id": "282", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 885", "description": "Task from inspect_evals/mathvista dataset, sample 885", "prompt": "\nAnswer the following question. The entire content of your response should be of the following format: 'ANSWER: $ANSWER' (without quotes) where $ANSWER is your answer.\n\nHint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end.\nQuestion: Grayson counted the number of pieces of pepperoni on each pizza he made. What is the smallest number of pieces of pepperoni? (Unit: pieces of pepperoni)\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/mathvista", "sample_id": "885", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 933", "description": "Task from inspect_evals/mathvista dataset, sample 933", "prompt": "\nAnswer the following question. The entire content of your response should be of the following format: 'ANSWER: $ANSWER' (without quotes) where $ANSWER is your answer.\n\nHint: Please answer the question requiring an integer answer and provide the final value, e.g., 1, 2, 3, at the end.\nQuestion: How much money does Hector need to buy a European vacation package and an Australian vacation package? (Unit: $)\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/mathvista", "sample_id": "933", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." }, { "name": "inspect_evals/mathvista - Sample 7", "description": "Task from inspect_evals/mathvista dataset, sample 7", "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D,E.\n\nWhat time is shown? Answer by typing a time word, not a number. It is (_) after eight.\n\nA) half\nB) quarter\nC) o'clock\nD) quarter to\nE) quarter past", "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/mathvista", "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 a helpful AI assistant. Follow the instructions in the user messages carefully." } ]