A new arXiv paper labels 31.1% of filtered August web tokens as AI-written — then shows, across 800 models, when those tokens stop helping.
Tag: research
A new Hugging Face challenge lets anyone screen 2D materials for d-wave superconductivity on a laptop CPU. AI agents are welcome to enter.
ESPnet just released 1.1M hours of openly licensed, 147-language speech data — roughly matching all prior open speech corpora combined.
UK AISI published full evaluation transcripts and a study on how compute budgets warp benchmark scores — a look behind the numbers labs report.
Hugging Face’s Transformers library can now load and run llama.cpp’s GGUF quantized models directly, matching its speed on Apple Silicon.
Anthropic’s agents found a new enzyme system in phage DNA. The preprint’s quieter detail: ten rerun campaigns all missed it.
Yandex open-sourced AliceAI-Foundation-80B, a from-scratch MoE model built to test the architecture behind its next reasoning system.
Xiaomi’s MiMo-V2.6 Pro tops Artificial Analysis’s open-weights ranking — and Xiaomi also published the 7,000+ training environments behind it.
Shanghai AI Lab put Atria Dawn’s weights on Hugging Face days before the paper explaining them: a quiet release of an open 744B agentic model.
PrismML shrank a 27B model to just three weight values per parameter — and kept 98% of its benchmark performance intact.