Core Foundation Model Research Intern

About the Role We are looking for an exceptional Research Intern to join our Core Foundation Model Research Team, working on the frontier of AGI-oriented large model technologies. You will contribute to next-generation foundation models by exploring sparse training, implicit reasoning, diffusion-based language modeling, self-improving systems, and long-term memory architectures. This internship is ideal for research-minded individuals who love pushing the boundaries of model capacity, reasoning, and emergent behaviors. What You’ll Do Research and prototype novel sparse training methods (e.g., Mixture-of-Experts, dynamic sparsity, conditional compute). Explore techniques for implicit reasoning and differentiable planning, such as MCTS-inspired modules inside LMs. Experiment with diffusion-based sequence modeling and new generative paradigms for language and multimodal tasks. Investigate self-learning loops, model bootstrapping, and unsupervised knowledge accumulation. Prototype architectures for long-term memory, retrieval-augmented generation, or lifelong learning. Collaborate closely with world-class researchers in LLMs, generative modeling, and cognitive architectures. Who You Are Must-Have Strong foundation in deep learning, especially large-scale transformer models. Hands-on experience with model training pipelines (PyTorch, JAX, DeepSpeed, or similar). Curiosity about model reasoning, emergent abilities, and self-improvement. Familiarity with reading and implementing cutting-edge research papers. Nice-to-Have Prior research or publications in sparse models, differentiable reasoning, or generative modeling. Knowledge of diffusion models, probabilistic generative methods, or planning algorithms. Familiarity with retrieval-augmented generation, vector DBs, or scalable memory systems. Experience contributing to open-source ML frameworks or custom model architectures Mindset You’re passionate about AGI-oriented research and unconstrained exploration You’re comfortable with rapid prototyping, experimental failures, and sharing insights. You thrive in interdisciplinary conversations and enjoy learning across model theory, cognition, and systems Why Join Us Work at the forefront of foundational model research with a real AGI ambition. Access massive compute, custom infra, and mentoring from world-class model builders. Collaborate with a diverse team spanning model algorithms, reasoning, memory systems, and training engineering. Publish, share, and test your ideas on real-world scale models.

暂无要求