Chinese long-context LLM benchmark V2 with harder NIAH variants, 10 repeats, and efficiency metrics for DeepSeek, Kimi, and Qwen.
-
Updated
May 2, 2026 - Jupyter Notebook
#
niah
Here are 3 public repositories matching this topic...
LLM-friendly encoding for random identifiers (hex, base64, UUID). Built for agents, RAG, and NIAH-style retrieval.
python-library rag pypi-package ai-tools agentic-rag token-optimization context-engineering agent-hooks niah
-
Updated
May 21, 2026 - Python
Honest measurement of 1M-token long-context benchmarks (RULER + LongBench v2 + NIAH) on Qwen2.5-7B-1M local vs GitHub Models cloud. All zero credit card, drift-checked, reproducible.
portfolio benchmark transformers llm long-context vllm bitsandbytes qwen github-models qwen2-5 niah consumer-laptop
-
Updated
Jun 1, 2026 - Jupyter Notebook
Improve this page
Add a description, image, and links to the niah topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the niah topic, visit your repo's landing page and select "manage topics."