Posts

All the articles I've posted.

• Enabling Flexible Multi-LLM Integration for Scalable Knowledge Aggregation

  Published: 4 Jun, 2025 |  at  11:27 AM

  

  92.94 🤔

  本文提出了一种动态整合框架，通过自适应选择网络和动态加权融合策略从多个LLM中聚合知识，显著提升性能并减少50%的知识干扰，同时保持计算效率。

• Local Mixtures of Experts: Essentially Free Test-Time Training via Model Merging

  Published: 23 May, 2025 |  at  11:15 AM

  

  92.79 🤔

  本文提出测试时模型合并（TTMM）方法，通过在训练时预训练大量专家模型并在测试时动态合并参数，以几乎无测试时开销的方式逼近测试时训练（TTT）的语言建模性能。

• Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories

  Published: 16 May, 2025 |  at  11:10 AM

  

  92.78 🤔

  This exploratory study evaluates GPT-4o's multilingual and multimodal performance on physics concept inventories, revealing strong results in English and text-based tasks but significant weaknesses in visual interpretation and non-Western languages, highlighting implications for equitable AI integration in education.

• Differentially Private Bilevel Optimization

  Published: 14 May, 2025 |  at  11:12 AM

  

  92.72 🤔

  This paper introduces the first differentially private first-order algorithms for bilevel optimization, ensuring privacy with theoretical convergence guarantees for hypergradient norms in both empirical and population settings while avoiding Hessian computations.

• Towards Revealing the Effectiveness of Small-Scale Fine-tuning in R1-style Reinforcement Learning

  Published: 28 May, 2025 |  at  11:25 AM

  

  92.52 🤔

  本文通过理论分析和Re-distillation技术，揭示了小规模SFT在R1风格RL中的效率瓶颈，并以极少样本（<1K）在K&K和MATH数据集上接近RL性能，显著提升了数据效率。