Long-context Question Answering on NovelQA
84.85AccuracyStateLM-14B-RL
Evaluation results
| Method | Links | |
|---|---|---|
| StateLM-14B-RLContext=32K2026.02 | 84.85 | |
| StateLM-8B-RLContext=32K2026.02 | 84.15 | |
| StateLM-14BContext=32K2026.02 | 84.15 | |
| StateLM-8BContext=32K2026.02 | 83.84 | |
| Qwen3-235B (w/ Pensieve)Context=256K2026.02 | 80.71 | |
| StateLM-4BContext=32K2026.02 | 79.57 | |
| RL-MemoryAgent-14BContext=32K2026.02 | 78.86 | |
| Qwen3-14BContext=128K2026.02 | 77.94 | |
| Qwen3-8BContext=128K2026.02 | 65.87 | |
| Qwen3-4BContext=128K2026.02 | 65.17 | |
| RL-MemoryAgent-7BContext=32K2026.02 | 60.24 | |
| ReadAgent-14BContext=32K2026.02 | 23.12 | |
| ReadAgent-8BContext=32K2026.02 | 16.38 |