Nishi Family › Compare › Frontier Radar
Where the 2025/26 research frontier is moving, and where Nishi should test, deepen, or research next.
| Capability axis | Nishi organ | Frontier momentum (2025/26) | Signal |
|---|---|---|---|
| RLHF / DPO preference alignment | absent | 120 mentions | OPPORTUNITY |
| Standardized eval harness | absent | 89 mentions | OPPORTUNITY |
| Reverse-mode autodiff training substrate | present | 80 mentions | COVERED |
| LoRA / parameter-efficient fine-tuning | absent | 61 mentions | OPPORTUNITY |
| Integer K-quant weights | present | 54 mentions | COVERED |
| Instruction fine-tuning (SFT) | absent | 46 mentions | OPPORTUNITY |
| Large-scale pretraining | absent | 32 mentions | OPPORTUNITY |
| Transformer forward (100 percent integer no-float) | present | 15 mentions | COVERED |
| Served chat API | present | 11 mentions | COVERED |
| GPU / CUDA training | absent | 10 mentions | OPPORTUNITY |
| Distributed tensor-parallel training | absent | 2 mentions | OPPORTUNITY |
| BPE tokenizer training | present | 0 mentions | COVERED |
| SIMD integer GEMM | present | 0 mentions | COVERED |
| Open-weights GGUF loading | present | 0 mentions | COVERED |
Opportunities 7 · covered 7 · dormant 0
frontier momentum is a mechanical keyword count over real 2024-2026 works; our presence is probed on disk. A discovery signal, not a ranking of importance. Zero JS, zero trackers.