π€ AI Agent7 items
Latest AI agent research papers, analysis, and improvement insights
semantic_scholarresearch
The Compaction Cliff in Long-Running AI Agent Memory
컨ν μ€νΈ μμΆ μ μμ κ·μΉμ΄ μμ€λλ λ¬Έμ λ₯Ό λ°κ²¬νκ³ Knowledge Triageλ‘ ν΄κ²°
semantic_scholarresearch
ExecRubrics: Executable Tool-Augmented Rubrics for Verifiable and Efficient Long-Form Evaluation
Tool κΈ°λ° LLM μμ΄μ νΈμ κ²°κ³Όλ¬Όμ νλ‘κ·Έλ맀ν±νκ² κ²μ¦ νκ°νλ λ°©μ
semantic_scholarresearch
Forgotten in Weights, Recovered by Tools: Agentic Tool Unlearning for LLM Agents
Unlearning νμλ λꡬλ₯Ό ν΅ν΄ μ 보 볡ꡬλλ λ¬Έμ μ κ³Ό λμ λ°©μ μ°κ΅¬
semantic_scholarresearch
From Language Models to Agentic AI: A Survey of Autonomous, Action-Enabled, and Collaborative LLM Agents
μμ΄μ νΈ μν€ν μ², λꡬ νμ©, νμ λ°©μκ³Ό νκ° νμ€μ μ’ ν©ν μ΅μ μλ² μ΄
semantic_scholarresearch
AI Watchdog: Agent Interfaces for Detecting and Defending Against Manipulative Dark Patterns in AI Conversations
AI λνμ 5κ°μ§ μ‘°μ ν¨ν΄μ μ€μκ° νμ§νλ λΈλΌμ°μ κΈ°λ° λ³΄νΈ μμ€ν
semantic_scholarresearch
Who Delegates to AI? Evidence from 53,000 Agent Configurations
53,000κ° μμ΄μ νΈ λΆμμΌλ‘ μ§μ λ³ AI λμ νν©κ³Ό μμ΄μ ν± μ±ν μ§μ λμΆ
semantic_scholarresearch
Designing a Robust LLM-Based Evaluation System for Agentic AI in Drug Discovery Through Human Alignment
μ½λ¬Ό λ°κ²¬ λλ©μΈμμ Tool κΈ°λ° μμ΄μ νΈμ μ±λ₯μ μΈκ° μ λ ¬μΌλ‘ νκ°