Time-controlled instruction-following on TiCo-Bench Text Short (10s–30s)
4.08MAE (s)TiCo
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TiCoModel Category=Proposed2026.03 | 4.08 | — | — | 21 | |
| TiCoModel Category=Proposed2026.03 | 4.46 | — | — | 22.7 | |
| Cascade (GPT)Model Category=Cascaded2026.03 | 4.83 | — | — | 23.5 | |
| Cascade (GPT)Model Category=Cascaded2026.03 | 5.92 | — | — | 28.8 | |
| Qwen2.5-Omni-7BModel Category=Open-sourced2026.03 | 7.73 | — | — | 44.9 | |
| Cascade (Qwen)Model Category=Cascaded2026.03 | 7.98 | — | — | 40.4 | |
| Mimo-AudioModel Category=Commercial2026.03 | 9.4 | — | — | 46.8 | |
| Cascade (Qwen)Model Category=Cascaded2026.03 | 10.38 | — | — | 51.2 | |
| LFM AudioModel Category=Commercial2026.03 | 12.43 | — | — | 60.4 | |
| Mimo-AudioModel Category=Commercial2026.03 | 12.66 | — | — | 62 | |
| Qwen2.5-Omni-7BModel Category=Open-sourced2026.03 | 13.74 | — | — | 79.1 | |
| GPT-audioModel Category=Commercial2026.03 | 14.33 | — | — | 73.6 | |
| GPT-audioModel Category=Commercial2026.03 | 18.75 | — | — | 97.1 | |
| LFM AudioModel Category=Commercial2026.03 | 18.84 | — | — | 101.1 | |
| Cascade (GPT)Backbone=GPT2026.03 | — | 2.86 | 4.18 | — | |
| Cascade (Qwen)Backbone=Qwen2026.03 | — | 2.32 | 3.35 | — | |
| GPT-audio2026.03 | — | 2.55 | 3.23 | — | |
| LFM Audio2026.03 | — | 1.93 | 2.16 | — | |
| Mimo-Audio2026.03 | — | 1.42 | 1.53 | — | |
| Qwen2.5-Omni-7BParameters=7B2026.03 | — | 2.3 | 3.32 | — | |
| TiCo (Proposed)2026.03 | — | 2.23 | 3.33 | — |