| [ |
| { |
| "name": "inspect_evals/docvqa - Sample 50311", |
| "description": "Task from inspect_evals/docvqa dataset, sample 50311", |
| "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\nWhat is the time of message?\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/docvqa", |
| "sample_id": "50311", |
| "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/docvqa - Sample 51240", |
| "description": "Task from inspect_evals/docvqa dataset, sample 51240", |
| "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\nWhat is the first heading?\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/docvqa", |
| "sample_id": "51240", |
| "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/docvqa - Sample 56032", |
| "description": "Task from inspect_evals/docvqa dataset, sample 56032", |
| "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\nWhat is the total number of shares transferred on 02/04/2004?\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/docvqa", |
| "sample_id": "56032", |
| "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/docvqa - Sample 43807", |
| "description": "Task from inspect_evals/docvqa dataset, sample 43807", |
| "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\nHow much is 'Year-to-Date Sales'?\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/docvqa", |
| "sample_id": "43807", |
| "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/docvqa - Sample 55115", |
| "description": "Task from inspect_evals/docvqa dataset, sample 55115", |
| "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\nWhat is the \u2018consultant costs\u2019 for \u20181st period (same as detailed budget)\u2019?\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/docvqa", |
| "sample_id": "55115", |
| "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/docvqa - Sample 62076", |
| "description": "Task from inspect_evals/docvqa dataset, sample 62076", |
| "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\nWhat is the location of Forty-Niners Award ceremony?\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/docvqa", |
| "sample_id": "62076", |
| "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/docvqa - Sample 43229", |
| "description": "Task from inspect_evals/docvqa dataset, sample 43229", |
| "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\n'Have business needs changed since original approval'?\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/docvqa", |
| "sample_id": "43229", |
| "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/docvqa - Sample 60390", |
| "description": "Task from inspect_evals/docvqa dataset, sample 60390", |
| "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\nWho is the message for?\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/docvqa", |
| "sample_id": "60390", |
| "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/docvqa - Sample 49391", |
| "description": "Task from inspect_evals/docvqa dataset, sample 49391", |
| "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\nWhat is the heading numbered 'IV' ?\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/docvqa", |
| "sample_id": "49391", |
| "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/docvqa - Sample 58416", |
| "description": "Task from inspect_evals/docvqa dataset, sample 58416", |
| "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\nWhat is the budget for the symposia?\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/docvqa", |
| "sample_id": "58416", |
| "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." |
| } |
| ] |
