Latency Evaluation on MeetingBank
0.4Latency (s)LLMLingua-2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLMLingua-2Compression Ratio=3x, Evaluation Protocol=Compression Overhead2024.03 | 0.4 | — | |
| LLMLingua-2Compression Ratio=5x, Evaluation Protocol=Compression Overhead2024.03 | 0.4 | — | |
| LLMLingua-2Compression Ratio=2x, Evaluation Protocol=Compression Overhead2024.03 | 0.5 | — | |
| LLMLinguaCompression Ratio=5x, Evaluation Protocol=Compression Overhead2024.03 | 1.5 | — | |
| LLMLinguaCompression Ratio=3x, Evaluation Protocol=Compression Overhead2024.03 | 2.1 | — | |
| LLMLinguaCompression Ratio=2x, Evaluation Protocol=Compression Overhead2024.03 | 2.9 | — | |
| End2End w/ LLMLingua-2Compression Ratio=5x, Evaluation Protocol=End-to-End System2024.03 | 5.2 | 2.9 | |
| End2End w/ LLMLingua-2Compression Ratio=3x, Evaluation Protocol=End-to-End System2024.03 | 7.5 | 2.1 | |
| End2End w/ LLMLingua-2Compression Ratio=2x, Evaluation Protocol=End-to-End System2024.03 | 9.4 | 1.6 | |
| End2End w/o CompressionCompression Ratio=1x, Evaluation Protocol=End-to-End System2024.03 | 14.9 | — | |
| Selective-ContextCompression Ratio=5x, Evaluation Protocol=Compression Overhead2024.03 | 15.5 | — | |
| Selective-ContextCompression Ratio=3x, Evaluation Protocol=Compression Overhead2024.03 | 15.6 | — | |
| Selective-ContextCompression Ratio=2x, Evaluation Protocol=Compression Overhead2024.03 | 15.9 | — |