Multi-Agent Image Integrity Audit for Antibody Validation
Automated image manipulation detection in publication figures is critical to biomedical research integrity. Build a multi-agent auditing workflow using AutoGen and Replit Agent to evaluate western blots and microscopy images for duplicate bands and splicing. Achieve greater than 90 percent detection accuracy on synthetic validation sets.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate an AutoGen multi-agent system combining computer vision pipelines and Replit Agent execution to detect image manipulation in biomedical literature.
How work is evaluated
Evaluates the accuracy of multi-agent image manipulation detection and audit report generation.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
valid_json_output
Ensures the agent outputs valid JSON schema matching audit specs.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
f1_score
F1 score of image manipulation anomaly detection. • target: 0.9 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Design an AutoGen conversational multi-agent system for automated figure inspection
Implement deep learning image forgery detection models using PyTorch and OpenCV
Automate code execution environments using Replit Agent tools
Generate standardized audit reports detailing integrity confidence scores
Reference links and supporting material
Synthetic and annotated collection of Western blot and microscopy images containing original panels and simulated splicing or duplication anomalies.
Challenge setup
Evaluates the accuracy of multi-agent image manipulation detection and audit report generation.
Challenge input
JSON array of image paths and metadata
AutoGen
Required framework for orchestrating conversational audit agents.
Replit Agent
Used for live code generation and isolated image metric computation.
Evaluated output
JSON array of bounding boxes, anomaly types, and overall verdict
- Ensures the agent outputs valid JSON schema matching audit specs.
- F1 score of image manipulation anomaly detection. • target: 0.9 • range: 0-1
- Benchmark: BioImageIntegrity-Bench
- F1 Score target: 0.9
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Replit Agent · Required
- Replit Agent · Required
- AutoGen · Optional
- Hunyuan · Optional
Evaluation contract
- valid_json_output · Weight 1
- f1_score · Weight 1
Metrics
- f1_score · Target 0.9
Benchmarks
- BioImageIntegrity-Bench
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON array of image paths and metadata
Output: JSON array of bounding boxes, anomaly types, and overall verdict
Execution harness: Python
Evaluation tools
- AutoGen · Multi-agent framework
- Replit Agent · Sandbox agent environment
Recipe freeze marker
No recipe freeze time is recorded.