Generative Recommendation on Scientific
3Recall@5Mamba2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Mamba2Throughput (Tput)=517, Speedup (Spd.)=1.22×2026.05 | 3 | 4.68 | 1.93 | 2.47 | |
| SID-MLPThroughput (Tput)=3,706, Speedup (Spd.)=8.74×2026.05 | 2.97 | 4.72 | 1.93 | 2.5 | |
| TIGERThroughput (Tput)=320, Speedup (Spd.)=0.75×2026.05 | 2.95 | 4.57 | 1.91 | 2.43 | |
| TIGER-kvThroughput (Tput)=424, Speedup (Spd.)=1.00×2026.05 | 2.95 | 4.57 | 1.91 | 2.43 | |
| AtSpeed-SThroughput (Tput)=286, Speedup (Spd.)=0.68×2026.05 | 2.95 | 4.57 | 1.91 | 2.43 | |
| SID-MLP++Throughput (Tput)=4,347, Speedup (Spd.)=10.25×2026.05 | 2.95 | 4.59 | 1.92 | 2.44 | |
| EARNThroughput (Tput)=822, Speedup (Spd.)=1.94×2026.05 | 2.88 | 4.54 | 1.86 | 2.39 | |
| GDNThroughput (Tput)=510, Speedup (Spd.)=1.20×2026.05 | 2.78 | 4.39 | 1.83 | 2.35 | |
| NEZHAThroughput (Tput)=3,082, Speedup (Spd.)=7.27×2026.05 | 2.54 | 4.02 | 1.64 | 2.12 | |
| AtSpeed-RThroughput (Tput)=94, Speedup (Spd.)=0.22×2026.05 | 2.16 | 3.2 | 1.44 | 1.78 |