.@SentientAGI recently announced that the world's largest artificial intelligence conference @NeurIPSConf has accepted four papers from the team covering different categories. After delving deeper, I found that this is a significant recognition of Sentient's technological innovations in the AI field. First, let me provide a brief introduction: NeurIPS (Conference on Neural Information Processing Systems) is one of the most influential academic conferences in the field of artificial intelligence and machine learning, established in 1987. Every year, it gathers top scientists and tech companies from around the world to showcase the latest breakthroughs in AI technology. From deep learning to large language models, many world-changing innovations have emerged here, which is why NeurIPS is also referred to as the “annual barometer” of the AI field. NeurIPS is known for its high academic standards and rigorous peer review, with an acceptance rate of about 20% for submissions over the years. Being accepted by the conference means that the paper has received high recognition from industry experts in terms of innovation, scientific merit, and potential impact. Sentient's recognized work due to its groundbreaking progress includes: ❶OML 1.0 (Main Track): Scalable LLM Fingerprinting Sentient has made a hundredfold improvement on traditional fingerprinting attempts for open models, injecting 24,576 persistent fingerprints. After fingerprint injection, the model's performance did not decline, meaning secure tracking can be achieved without compromising reasoning or generation capabilities. This can be considered the cornerstone of open-source AGI; only by ensuring that creators' interests are not harmed can the entire community develop sustainably. ❷LiveCodeBenchPro (Dataset and Benchmark Track) A custom benchmark designed for evaluating coding capabilities, accurately measuring the model's overall performance in real programming tasks. Sentient created a model scaled down by 10 times using 20% of the data, validating that high-quality data and task-oriented training can enable small models to achieve the capabilities of large models. ❸MindGames Arena (Competition Track) This track mainly publishes papers and proposals for AI competitions. Sentient proposed that NeurIPS select and host an AI competition aimed at enhancing agents' capabilities through social games. This can be understood as an exploration of the next stage of AI, focusing more on AI self-optimization and social learning capabilities. ❹OML Lock-LLM (Workshops and Security Technologies) An extension of OML technology that allows for verifiable, encrypted mandatory control of the model under white-box access conditions. It strikes a balance between transparency and security for open models—meaning that even if others can access the model parameters, they cannot modify or misuse them without authorization....