Error Attribution on GSM8k Mid Dense
75Removal RateVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=MACNET2026.05 | 75 | |
| VanillaScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=DyLAN2026.05 | 66 | |
| Two-WayScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=DyLAN2026.05 | 63 | |
| Two-WayScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 62 | |
| Two-WayScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=MACNET2026.05 | 60 | |
| Two-WayScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 59 | |
| Two-WayScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 59 | |
| Two-WayScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 59 | |
| Left FilteringScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 53 | |
| Left FilteringScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 47 | |
| Right FilteringScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 45 | |
| Right FilteringScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=DyLAN2026.05 | 44 | |
| Right FilteringScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 43 | |
| Right FilteringScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 43 | |
| CRSVPScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=MACNET2026.05 | 38 | |
| Left FilteringScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 36 | |
| CRSVPScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=DyLAN2026.05 | 36 | |
| Left FilteringScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=MACNET2026.05 | 35 | |
| Right FilteringScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 33 | |
| Right FilteringScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=MACNET2026.05 | 33 | |
| Left FilteringScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 31 | |
| Left FilteringScoring Function Type=Fine-tuned, Backbone=Qwen3-1.7B, Agentic Framework=DyLAN2026.05 | 31 | |
| CRSVPScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 29 | |
| VanillaScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 27 | |
| VanillaScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 27 | |
| CRSVPScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 27 | |
| CRSVPScoring Function Type=Role-prompted, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 26 | |
| CRSVPScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 21 | |
| VanillaScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=DyLAN2026.05 | 18 | |
| VanillaScoring Function Type=Naive LLM, Backbone=gpt-4o-mini, Agentic Framework=MACNET2026.05 | 15 |