Llm Powered Abstractive Summarization With Semantic Compression

via “content summarization and extraction”

text-generation model by undefined. 94,68,562 downloads.

Unique: Instruction-tuned abstractive summarization using full 128K context window to process entire documents without chunking; learns summarization patterns from training data rather than using extractive algorithms, enabling flexible output formats and style adaptation

vs others: Handles longer documents than Mistral-7B (smaller context) and provides more flexible summarization than rule-based extractive tools; comparable to GPT-3.5 on quality but with local deployment and no API costs