-
Notifications
You must be signed in to change notification settings - Fork 0
Open
Description
Description:
We need to extract the fraction of feeding predators from PDFs classified as useful. Current exploration using the LLM model (Ollama phi3:3.8b) has shown difficulties in reliably extracting vague or inconsistently formatted data in a RAG-based workflow.
Goals:
- Establish a workflow that efficiently handles vague or inconsistent data from predator diet survey PDFs.
Notes / Blockers:
- Extraction is challenging due to inconsistent formatting and vague data references.
- Need to determine if additional preprocessing, prompts, or model adjustments are required for reliable extraction.
Acceptance Criteria:
- Extraction model reliably outputs fraction of feeding predators with minimal errors.
- Workflow is reproducible and documented for team use.
Metadata
Metadata
Assignees
Labels
No labels