Southern University of Science and Technology and IDEA Research Institute Release SUS-Chat-34B Bilingual Model

The Southern University of Science and Technology, in collaboration with the IDEA Institute's CCNL Center, has released the SUS-Chat-34B bilingual model, which boasts a parameter scale of 34 billion. This model excels in both Chinese and English tasks, surpassing other models of the same parameter size. SUS-Chat-34B is fine-tuned on the Yi-34B pre-trained model, utilizing millions of high-quality, multilingual instruction data for training. The model features extensive complex instruction-following data, robust performance across general tasks, and an extended context window with exceptional multi-turn dialogue capabilities.

Xiaohongshu Officially Announces Organizational Restructuring: Conan Appointed as President, Establishes AI Primary Department Dots and Overseas Division Rednote

On April 30, Xiaohongshu announced an organizational upgrade, integrating community, e-commerce, commercialization, and technology systems, and establishing a new AI first-level department 'Dots', elevating AI to the highest strategic priority for transformation from tool to core productivity. Ke Nan (Ding Ling) was appointed president, overseeing core business and reporting directly to CEO Xing Shi.....

OpenAI System Prompt Leaked, New Model GPT-5.5 Strictly Prohibits Discussion of Goblins

OpenAI's latest open-source Codex CLI accidentally exposed GPT-5.5's system prompts, including a mysterious instruction prohibiting discussions of fantasy creatures like 'goblins' and 'elves' in conversations. The 3,500-word base instruction set mandates that the model avoid these topics unless user queries have absolutely clear relevance, aiming to prevent AI from falling into specific hallucinations.....

OpenAI New Model System Instructions Leaked, GPT-5.5 Is Now Banned from Talking about Goblins?

OpenAI's latest open-source Codex CLI reveals partial underlying logic of GPT-5.5. According to Ars Technica, its 3,500+ word system prompt includes a rare instruction: strictly prohibit mentioning specific creatures like 'goblin' without explicit relevance, unless absolutely necessary for the query.....

1% Performance Improvement Is a Thing of the Past? CVPR 2026 Reveals a Paradigm Shift in Visual Intelligence

Computer vision is shifting from 'seeing the world' to 'understanding and interaction.' As perception approaches human limits, the marginal gains from pursuing accuracy alone diminish. CVPR2026 marks a shift in focus: vision becomes a medium for reasoning, decision-making, and interaction, moving away from 'blind reasoning' toward adaptive and implicit pathways, such as multimodal models using 'chain-of-thought' for logic.....

Southern University of Science and Technology and IDEA Research Institute Release SUS-Chat-34B Bilingual Model

Related Recommendations

Microsoft's Revenue Surpasses $82 Billion in Third Fiscal Quarter, AI Revenue Soars 123% and Becomes a Core Growth Driver

Xiaohongshu Officially Announces Organizational Restructuring: Conan Appointed as President, Establishes AI Primary Department Dots and Overseas Division Rednote

OpenAI System Prompt Leaked, New Model GPT-5.5 Strictly Prohibits Discussion of Goblins

OpenAI New Model System Instructions Leaked, GPT-5.5 Is Now Banned from Talking about Goblins?

1% Performance Improvement Is a Thing of the Past? CVPR 2026 Reveals a Paradigm Shift in Visual Intelligence