Back to evaluations
Public evaluation

Disinformation Analysis

Evaluation will focus on the system's ability to accurately detect and explain disinformation, its resilience to subtle manipulations, and the efficiency of its agentic workflow.

Evaluation type
task based
Challenge
Build a Factual Integrity Agent System
Difficulty
Advanced
Rigor
Unspecified
Open linked challenge

Evaluation overview

How the linked challenge is judged: tasks, benchmarks, and criteria count.

Tasks
2
Benchmarks
0
Criteria
0

Task templates

Inputs and expected outputs.

Task 1

Disinformation Analysis

Analyze a provided text snippet containing a mix of factual and subtly manipulated information, identifying disinformation and explaining why.

Input format

{ "text": "string", "topic": "string" }

Output format

{ "disinformation_detected": "boolean", "identified_claims": [ { "claim": "string", "is_disinformation": "boolean", "evidence": "string", "explanation": "string" } ], "confidence": "float" }

Task 2

Extended Thinking Efficiency

Measure the number of reasoning steps and tool calls for a given complex verification task, demonstrating adaptive budget utilization.

Input format

{ "query": "string", "max_iterations": "int" }

Output format

{ "final_answer": "string", "reasoning_steps": "int", "tool_calls": ["string"] }