TableGPT2-7B is a large-scale decoder optimized for data-intensive tasks, specializing in the interpretation and analysis of tabular data. huggingface.co/tablegpt/Table…

After enabling the WebGPU feature in Chrome on my Android and testing it with the Qwen 1.5B model, I fully understand why deploying LLMs on mobile devices is challenging. Even with a 1.5B webassembly model, the performance is still too slow. (Local .so might be more efficient

哇！铁山靠！youtu.be/rDolt3jJRsM?t=…

Blame the gerbils lrb.co.uk/the-paper/v46/…

論文引文也要講性別，瑟瑟發抖 "Edited only to reduce gender-specific references." DEFENSIVE COMMUNICATION reagle.org/joseph/2010/co…

Good job! "Bypassing regulatory locks, Faraday cages and upgrading your hearing" lagrangepoint.substack.com/p/airpods-hear…

Something weird is happening with LLMs and chess dynomight.substack.com/p/chess

看 AI 谈恋爱神似逛弱智吧 ...

Bayesian Neural Networks cs.toronto.edu/~duvenaud/dist…

tej.as/blog/how-to-gr…

divisbyzero.com/2024/11/30/mak…

Aha! ai.google.dev/gemini-api/doc… is down. Error: Server Error The server encountered an error and could not complete your request.Please try again in 30 seconds.

沉浸式翻译里面（接 ollama） qwen3 0.6b 的速度比 qwen2.5 3b 还要慢，我不理解

日常重度 gemini api 使用者：不能同意更多 venki.dev/notes/google-g…

Reason-ModernColBERT is a late interaction model trained on the reasonir-hq dataset. It achieves extremely competitive performance on the BRIGHT benchmark... huggingface.co/lightonai/Reas…

velvetshark.com/ai-company-log…

GitHub free LLM API token "How much it costs: Free tier for all personal accounts and OSS orgs; metered paid tier unlocks higher throughput and larger context windows." docs.github.com/en/github-mode…

vercel is down 20251020

MTEB Leaderboard is broken 20251115

I found a special google ui google.com/search?q=&udm=…