Semantic Window Slicing
Cross-source consensus on Semantic Window Slicing from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Benefits
Preparation
Comparisons
Highlighted claims
- Inputs used plus-or-minus five-line windows around prioritized clinical headers such as FINAL DIAGNOSIS. — Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction
- Full text and semantic slicing performed similarly, while random windows substantially degraded performance. — Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction
- Each model input followed a four-block schema containing Field, Classes, Report, and Hints. — Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction
- Semantic slicing is framed as a deployment optimization that preserves accuracy while reducing sequence length and self-attention cost. — Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction
- Hints guided model attention but did not constrain outputs with rules. — Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction