Long-Context Language Modeling on LongBench 1.0 (test)
32.6NrtvQAFULL
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FULLModel=Qwen3-32B, Budget=N/A2026.04 | 32.6 | 45.5 | 50.4 | 59.6 | 56 | 40.1 | 33.2 | 23.9 | 25.2 | 72 | 70.8 | 37.1 | 18 | 100 | 12.4 | 18.4 | 43.4 | |
| ICEModel=Qwen3-32B, Budget=2562026.04 | 32.2 | 44.1 | 51.9 | 60.2 | 55.1 | 38.9 | 32.4 | 24.3 | 24.7 | 71.5 | 70.7 | 37.4 | 16.5 | 100 | 11.8 | 18 | 43.1 | |
| ICEModel=Qwen3-32B, Budget=1282026.04 | 32.1 | 44.8 | 53.3 | 61.6 | 54.6 | 38.5 | 31 | 23.6 | 24.8 | 72 | 70.3 | 36.2 | 15.5 | 100 | 10.7 | 17.3 | 42.6 | |
| ICEModel=Qwen3-32B, Budget=642026.04 | 29.5 | 43.9 | 50.9 | 61.4 | 55.8 | 38.8 | 30.8 | 23.8 | 24.3 | 71 | 70 | 35.7 | 15 | 97 | 10.5 | 17.3 | 42.2 | |
| FULLModel=LongChat-7B-v1.5, Budget=N/A2026.04 | 20.8 | 29.4 | 43.1 | 33 | 24.4 | 14.7 | 30.8 | 22.8 | 26.7 | 66.5 | 84 | 22.5 | 0 | 30.5 | 54.7 | 59.2 | 35.2 | |
| ICEModel=LongChat-7B-v1.5, Budget=2562026.04 | 20.4 | 29.5 | 43 | 34.6 | 23.7 | 14.2 | 29.8 | — | 26.1 | 66.5 | 84.7 | 22.9 | 0 | 28.5 | 53.6 | 59 | 35 | |
| ICEModel=LongChat-7B-v1.5, Budget=1282026.04 | 19.9 | 27.7 | 40.9 | 33.9 | 24.2 | 14.4 | 28.6 | 21.8 | 25.9 | 66.5 | 84.2 | 22.3 | 1.5 | 27.3 | 52.6 | 57.7 | 34.3 | |
| ICEModel=LongChat-7B-v1.5, Budget=642026.04 | 18.5 | 26.9 | 40.6 | 34.2 | 22.7 | 14 | 29.3 | 20.9 | 25.6 | 66.5 | 84.3 | 21.9 | 1.5 | 26.1 | 52.5 | 56.8 | 33.9 |