| Paper | Code | Accuracy | ModelName | ReleaseDate |
|---|---|---|---|---|
| Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding | ✓ Link | 78.5 | Confident Decoding (Qwen3.5-122B-A10B) | 2026-06-20 |
| LLaDA2.1: Speeding Up Text Diffusion via Token Editing | ✓ Link | 54.1 | LLaDA2.1-flash (100B) | 2026-02-09 |
| CoTranslator: Reverse to Forward Chain-of-Thought Translator for Reasoning Distillation | 50.99 | CoTranslator (Qwen3-4B-Thinking-2507) | 2026-10-14 | |
| LLaDA2.0: Scaling Up Diffusion Language Models to 100B | ✓ Link | 50.3 | LLaDA2.0-flash | 2025-12-10 |
| Semifactual Credit-Augmented Policy Optimization | ✓ Link | 39.42 | SCAPO (Qwen3-4B-Base) | 2026-09-30 |