| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| Needle-in-a-Haystack (test) | Accuracy100 | 56 | 3mo ago | ||
| NIAH | Accuracy100 | 50 | 24d ago | ||
| RULER | Retrieval Accuracy (8K)98.14 | 44 | 2mo ago | ||
| MLDR | moBERTo (orig. tok.) | nDCG@1062.86 | 36 | 1mo ago | |
| RULER 16K | HiLS-Attn HoPE-Q-Cal | Score98.67 | 34 | 18d ago | |
| Needle-in-the-Haystack 10k-context | Quest | Accuracy100 | 30 | 2mo ago | |
| BRIGHT StackExchange | Claude Sonnet 4.5 | Biology Score62.3 | 29 | 2mo ago | |
| Needle-in-a-Haystack | Retrieval Accuracy100 | 29 | 1mo ago | ||
| S-NIAH | Latency (s)10.5 | 27 | 4mo ago | ||
| RULER 128K | HiLS-Attn HoPE-Q-Cal | Score94.67 | 23 | 18d ago | |
| NIAH Single 3 | Looped Hybrid (GDN+DSA) | Accuracy (1024)100 | 22 | 1mo ago | |
| NIAH 128k | Single Score24.4 | 20 | 3mo ago | ||
| NIAH 64k | Single Score49.3 | 20 | 3mo ago | ||
| Lost-in-the-Middle 30-passage contexts | PRISM-∆ | Average Exact Match62.57 | 20 | 4mo ago | |
| NIAH multivalue | FLy | Speedup4.1 | 20 | 4mo ago | |
| RULER 64K context | WindowedManifoldKV | Accuracy84.3 | 19 | 2mo ago | |
| RULER 64K | HiLS-Attn HoPE-Q-Cal | Score97.33 | 17 | 18d ago | |
| RULER 32k context (test) | Ministral-3-8B-Instruct-2512-BF16 | Accuracy91.7 | 16 | 2mo ago | |
| RULER 4k context (test) | Llama-3.1-8B-Instruct | Accuracy95 | 16 | 2mo ago | |
| RULER 8K | Olmo3 512swa | Score99.67 | 15 | 18d ago | |
| RULER 32K | Hybrid-KSA | RULER 32K Retrieval Score86.65 | 15 | 1mo ago | |
| RULER | Accuracy80.1 | 14 | 3mo ago | ||
| NIAH-Multi | Kimi-K2 | Accuracy100 | 13 | 4mo ago | |
| RULER | Qwen3-1.7B | Retrieval Accuracy (4K Context)100 | 13 | 1mo ago | |
| Needle-in-a-Haystack 64K-token context | Success Rate98.7 | 12 | 17d ago |