Table Reasoning Accuracy on TableBench
63.48AccuracySERO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SEROBackbone=Gemini-2.5-flash-lite2026.05 | 63.48 | — | |
| Static Role OrchestrationBackbone=Gemini-2.5-flash-lite2026.05 | 61.6 | — | |
| Static DAG MASBackbone=Gemini-2.5-flash-lite2026.05 | 60.68 | — | |
| Random Role EvolutionBackbone=Gemini-2.5-flash-lite2026.05 | 60.64 | — | |
| CoTBackbone=Gemini-2.5-flash-lite2026.05 | 59 | — | |
| SC@3Backbone=Gemini-2.5-flash-lite2026.05 | 58.5 | — | |
| WorkflowBackbone=Gemini-2.5-flash-lite2026.05 | 57.91 | — | |
| 14B OnlyBackbone=Qwen3-Instruct, Extra Tokens=–2026.05 | 54.85 | 42.4 | |
| SEROBackbone=GPT-4o-mini2026.05 | 54.06 | — | |
| 14B OnlyBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=–2026.05 | 53.65 | 40.5 | |
| Static Role OrchestrationBackbone=GPT-4o-mini2026.05 | 52.43 | — | |
| Static DAG MASBackbone=GPT-4o-mini2026.05 | 52.3 | — | |
| EcoTabBackbone=Qwen3-Instruct, Extra Tokens=✗2026.05 | 51.77 | 28.1 | |
| Random Role EvolutionBackbone=GPT-4o-mini2026.05 | 51.76 | — | |
| GlimpRouterBackbone=Qwen3-Instruct, Extra Tokens=✓2026.05 | 50.52 | 30.1 | |
| EcoTabBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✗2026.05 | 50.41 | 27.4 | |
| STEERBackbone=Qwen3-Instruct, Extra Tokens=✗2026.05 | 50.33 | 29.4 | |
| SpecReasonBackbone=Qwen3-Instruct, Extra Tokens=✓2026.05 | 50.02 | 29.9 | |
| SpecCoTBackbone=Qwen3-Instruct, Extra Tokens=✓2026.05 | 49.87 | 29.7 | |
| GlimpRouterBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✓2026.05 | 49.54 | 28.1 | |
| SEROBackbone=Qwen3-8b2026.05 | 49.37 | — | |
| STEERBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✗2026.05 | 49.22 | 29.1 | |
| RSDBackbone=Qwen3-Instruct, Extra Tokens=✓2026.05 | 49.21 | 30.4 | |
| WorkflowBackbone=GPT-4o-mini2026.05 | 49.08 | — | |
| SpecReasonBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✓2026.05 | 48.89 | 28.4 | |
| RSDBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✓2026.05 | 48.81 | 28.6 | |
| SpecCoTBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✓2026.05 | 48.74 | 28.3 | |
| Random Role EvolutionBackbone=Qwen3-8b2026.05 | 48.32 | — | |
| SC@3Backbone=GPT-4o-mini2026.05 | 47.99 | — | |
| Static Role OrchestrationBackbone=Qwen3-8b2026.05 | 47.95 | — | |
| CoTBackbone=GPT-4o-mini2026.05 | 47.24 | — | |
| RandomBackbone=Qwen3-Instruct, Extra Tokens=✗2026.05 | 46.12 | 36.7 | |
| WorkflowBackbone=Qwen3-8b2026.05 | 44.43 | — | |
| Static DAG MASBackbone=Qwen3-8b2026.05 | 44.05 | — | |
| RandomBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=✗2026.05 | 44.03 | 36.4 | |
| 1.7B OnlyBackbone=Qwen3-Instruct, Extra Tokens=–2026.05 | 42.55 | 6.12 | |
| 1.7B OnlyBackbone=DeepSeek-R1-Distill-Qwen, Extra Tokens=–2026.05 | 42.55 | 6.12 | |
| CoTBackbone=Qwen3-8b2026.05 | 29.44 | — | |
| SC@3Backbone=Qwen3-8b2026.05 | 29.27 | — |