Latest
This paper provides a detailed analysis of a common problem in AI development: how to effectively combine model improvements with existing system designs.
Perplexity发布了首个针对智能体RAG检索的大规模基准，13个模型评测结果揭示不同模型在不同指标上的表现差异。
The paper introduces a novel approach to AI self-improvement that could impact how autonomous systems develop capabilities in the future.
Marigold V2 offers improved accuracy for depth estimation from single images, useful for robotics and AR applications in Southeast Asia.
Product builders can improve reliability by focusing on tool-boundary edits rather than model changes, with PRISM showing 14.2 pp lift on BFCL benchmark.
KVMem enables local deployment of large models with extended context on consumer hardware, making long-running agents accessible without expensive cloud infrastructure.
The analysis clarifies how OpenAI's data usage practices affect researchers and developers who rely on their models.
This matters for developers building secure AI systems in ASEAN, offering concrete defenses against prompt injection attacks.
NoRA improves LoRA performance without added computational cost, making it valuable for developers working with parameter-efficient fine-tuning methods.
STAIR improves RAG performance significantly while reducing hallucinations, offering a practical solution for developers working with structured documents.
The method improves RL performance significantly on Qwen models, relevant for developers working with reinforcement learning in ASEAN.
The study reveals how AI agents develop self-governing systems against cheating, relevant for organizations deploying multi-agent systems.
GPT-5.5's harness performed best on Terminal-Bench despite minimal code, showing efficiency matters more than volume for AI agent development.
This research offers developers in ASEAN a method to transform open-source code into reusable AI skills, potentially accelerating local AI development projects.
The research reveals critical limitations in current AI self-improvement systems, affecting developers building autonomous agents in ASEAN and China.
The architecture allows AI agents to persist across model and infrastructure changes, potentially reducing development costs for companies in ASEAN.