Multi-turn reasoning on MT-Add
1.001Math LiC ScoreRLSTA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RLSTABackbone=Qwen2.5-7B-Instruct2026.03 | 1.001 | 0.688 | 0.691 | 0.596 | 0.744 | |
| RLSTA (+ Abstain)Backbone=Qwen2.5-7B-Instruct, Condition=Explicit instruction to abstain2026.03 | 0.988 | 0.702 | 0.762 | 0.642 | 0.773 | |
| RLAARBackbone=Qwen2.5-7B-Instruct2026.03 | 0.95 | 0.53 | 0.873 | 0.702 | 0.764 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.03 | 0.728 | 0.365 | 0.564 | 0.525 | 0.546 |