Science Question Answering on GPQA (OTPS, Accuracy, Drop)
180.31OTPSXShare
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| XSharewarm-up k0=1, batch budget ml=0, per-request budget mr=4, Batch Size=4, Speculation Length=32026.02 | 180.31 | 74.75 | -1.51 | |
| XSharewarm-up k0=1, batch budget ml=0, per-request budget mr=5, Batch Size=4, Speculation Length=32026.02 | 172.8 | 77.27 | 1.01 | |
| XSharewarm-up k0=2, batch budget ml=0, per-request budget mr=4, Batch Size=4, Speculation Length=32026.02 | 172.58 | 77.27 | 1.01 | |
| XSharewarm-up k0=0, batch budget ml=4, per-request budget mr=16, Batch Size=4, Speculation Length=32026.02 | 171.55 | 70.71 | -5.55 | |
| XSharewarm-up k0=0, batch budget ml=0, per-request budget mr=8, Batch Size=4, Speculation Length=32026.02 | 170.01 | 76.77 | 0.51 | |
| XSharewarm-up k0=1, batch budget ml=24, per-request budget mr=0, Batch Size=4, Speculation Length=32026.02 | 165.54 | 73.23 | -3.03 | |
| XSharewarm-up k0=2, batch budget ml=10, per-request budget mr=0, Batch Size=4, Speculation Length=32026.02 | 159.3 | 78.28 | 2.02 | |
| XSharewarm-up k0=1, batch budget ml=32, per-request budget mr=0, Batch Size=4, Speculation Length=32026.02 | 158.55 | 78.79 | 2.53 | |
| GPT-OSS 120B BaselineBatch Size=4, Speculation Length=32026.02 | 154.65 | 76.26 | — |