Question Answering on Long-context QA
82.7Task SuccessCVT-RL
Evaluation Results
| Method | Links | |
|---|---|---|
| CVT-RLVerifier=yes, Constraints=yes, CF rollouts=semantic, Belief=yes, Compute=3.92026.06 | 82.7 | |
| Information-matched CF processVerifier=yes, Constraints=yes, CF rollouts=semantic, Belief=yes, Compute=3.92026.06 | 78.4 | |
| Q-RAGVerifier=yes, Constraints=no, CF rollouts=no, Belief=no, Compute=1.42026.06 | 75.4 | |
| Compute-matched non-causalVerifier=yes, Constraints=yes, CF rollouts=random extra, Belief=yes, Compute=3.82026.06 | 75.2 | |
| LongRLVRVerifier=yes, Constraints=no, CF rollouts=no, Belief=no, Compute=1.22026.06 | 73.6 | |
| Constrained process RLVerifier=yes, Constraints=yes, CF rollouts=no, Belief=yes, Compute=1.52026.06 | 73.1 | |
| RLVMRVerifier=yes, Constraints=no, CF rollouts=no, Belief=no, Compute=1.32026.06 | 70.2 | |
| TROLL-styleVerifier=yes, Constraints=no, CF rollouts=no, Belief=no, Compute=1.12026.06 | 67 | |
| PPO-RLVRVerifier=yes, Constraints=no, CF rollouts=no, Belief=no, Compute=1.12026.06 | 63.2 | |
| SFTVerifier=no, Constraints=no, CF rollouts=no, Belief=no, Compute=1.02026.06 | 51.4 |