| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| OpenWebText | SDTT | Perplexity3.18 | 187 | 18d ago | |
| DomainBench | SYTTA-16 | BLEU (Agriculture)71.37 | 144 | 4mo ago | |
| WikiText-2 | Perplexity4.88 | 138 | 1mo ago | ||
| LM1B (test) | UDLM | Entropy2.46 | 90 | 1mo ago | |
| NoveltyBench | Diversity10 | 81 | 2mo ago | ||
| CNN/Daily Mail (test) | DEL | COH Score83 | 64 | 4mo ago | |
| GSM8K | LoRA + PCGrad | Accuracy86.4 | 63 | 1mo ago | |
| OpenWebText | IDLM-MDLM | Gen PPL11.312 | 54 | 1mo ago | |
| DFM 1024 samples masked backbone | TR-CIE | GPT-2 Perplexity135.92 | 42 | 1mo ago | |
| Medical Chatbot | ASR100 | 42 | 4mo ago | ||
| OWT | DFM (ESD) | GPT2 Perplexity5.33 | 41 | 3mo ago | |
| LM1B | DFM (ESD) | Perplexity (PPL)68.11 | 39 | 18d ago | |
| 5 Generation tasks | POP | Accuracy57.96 | 36 | 4mo ago | |
| Masked DFM backbone 1024 samples | θ-Trapezoidal | Unigram Entropy7.61 | 35 | 1mo ago | |
| Text Generation | PPL11.9 | 33 | 4mo ago | ||
| Text model inference M4 Max | vllm-mlx | Throughput (tok/s)525.5 | 31 | 4mo ago | |
| OpenWebText | Parallel (Ours) | Generative Perplexity22.099 | 30 | 23d ago | |
| 1024 samples | TR-CIE | Perplexity (GPT-2)104.44 | 30 | 1mo ago | |
| uniform DFM backbone 1024 samples | TR-CIE | LLaMA-3 Perplexity95.56 | 30 | 1mo ago | |
| MSCOCO | SARE | BLEU-157.2 | 26 | 4mo ago | |
| Math-500 | JetFlow | Throughput (TPS)1,094.6 | 25 | 1mo ago | |
| IFEval | Llama 3.1 8B Instruct | Accuracy74.49 | 23 | 1mo ago | |
| Wikitext-103 | Refined by Gemma3 27B | Perplexity32.88 | 23 | 3mo ago | |
| Spec-Bench Overall | SpecBound | SD Score2.33 | 21 | 3mo ago | |
| MMLU (test) | TAP | BS Score57.83 | 20 | 2mo ago |
