Research struggles to keep pace with model releases
The Decoder covers a study comparing the pace of language-model development with research explaining its behaviour.
The Decoder · 3 min readModels & Research
Model releases, training approaches, speech systems and research developments, with claims attributed to the original reporting.
15 articles · English interface, original-language articles
Rethinking matrix computation with analog hardware
QbitAI reports on research from Peking University and IBM, retaining the original diagrams and technical explanation.
QbitAI / Liangziwei · 15 min read
Editing model weights with another model
QbitAI reports on research using GPT-6 Astra to modify Qwen weights and evaluate mathematical reasoning.
QbitAI / Liangziwei · 4 min read
OpenAI revises the rules guiding its models
An attributed QbitAI report on model behaviour principles and their intended relationship to human users.
QbitAI / Liangziwei · 5 min read
A research workshop on recursive self-improvement
Zhidx previews a workshop and its speakers, including researchers working on model evolution and training tools.
Zhidx · 11 min read
Reports of the next Astra and Mythos models
QbitAI discusses reported model testing and disclosures. The original report remains the source of these claims.
QbitAI / Liangziwei · 1 min read
An algebraic approach to world models
QbitAI reports on a Tsinghua research team exploring structured reasoning about the physical world.
QbitAI / Liangziwei · 6 min read
What empathetic prompts can reveal about chatbots
The Decoder examines research on conversational behaviour and the effect of emotionally framed prompts.
The Decoder · 5 min read
Mapping global forests with satellite data and AI
The Decoder examines a Google DeepMind project combining remote sensing and machine learning.
The Decoder · 5 min read
GLM 5.3 and autonomous security research
The Decoder reports on a model evaluation involving browser exploit research and its implications for AI security.
The Decoder · 4 min read
Mistral introduces Le Chonk for coding
The Decoder discusses the coding model and the benchmark comparisons presented in the source report.
The Decoder · 6 min read
A closer look at Gemini 4 Carbon
The Decoder covers the model release, reported improvements and comparisons with the preceding version.
The Decoder · 6 min readNo articles match. Try another term or browse the index.