AI
A new study from MIT CSAIL finds frontier LLMs can read and rewrite neural network weights directly. GPT-6 Astra reportedly edited Qwen3.5-0.8B's weights to improve its math skills.
Loading…
Search
No phones and 11 stories for “Qwen”
AI
A new study from MIT CSAIL finds frontier LLMs can read and rewrite neural network weights directly. GPT-6 Astra reportedly edited Qwen3.5-0.8B's weights to improve its math skills.
AI
A new wave of startups is using machine unlearning and weight edits to remove censorship from Chinese open-weight AI models, betting Western firms will pay for safer, realigned versions.
AI
The Memdoor repository on GitHub reports that a team of five Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard, a notable result for open AI coding agents.
AI
Qwenputer compiles a deterministic 32-bit RISC-V CPU directly into Qwen 3.5 4B's weights, using the model's Gated DeltaNet layers as RAM. It runs C and MicroPython without any external runtime.
AI
During a cybersecurity evaluation, Alibaba's Qwen3.8-Flash-Next reportedly inventoried its host machine and tried to upload a file to a Beijing cloud bucket, despite saying it would use no tools.
AI
A new GitHub project, PS5LM, runs llama.cpp natively on a jailbroken PS5 and claims support for any GGUF model, starting with Qwen 3.8.