TechPulseTelegram
← PrevMon, Jun 29Next β†’

πŸ€– AI Agent7 items

Latest AI agent research papers, analysis, and improvement insights

semantic_scholarresearch

Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies

자체 μ§„ν™”ν•˜λŠ” LLM μ—μ΄μ „νŠΈμ˜ λ³΄μ•ˆ μœ„ν˜‘μ„ μ²΄κ³„μ μœΌλ‘œ λΆ„μ„ν•˜κ³  곡격 ν‘œλ©΄μ„ λΆ„ν•΄

semantic_scholarresearch

Semantic Early-Stopping for Iterative LLM Agent Loops

의미둠적 λ³€ν™” 기반 μ‘°κΈ° μ’…λ£Œλ‘œ μ—μ΄μ „νŠΈ 루프 토큰 νš¨μœ¨μ„± μ΅œμ ν™”

semantic_scholarresearch

Securing LLM-Agent Long-Term Memory Against Poisoning: Non-Malleable, Origin-Bound Authority with Machine-Checked Guarantees

λ©”λͺ¨λ¦¬ 쀑독 κ³΅κ²©μœΌλ‘œλΆ€ν„° LLM μ—μ΄μ „νŠΈλ₯Ό λ³΄ν˜Έν•˜λŠ” 좜처 기반 λ³΄μ•ˆ λ©”μ»€λ‹ˆμ¦˜

semantic_scholarresearch

Emergent Relational Order in LLM Agent Societies: From Collective Affect to Authority Stratification

닀쀑 μ—μ΄μ „νŠΈ ν”„λ ˆμž„μ›Œν¬μ—μ„œ 자발적 μ‚¬νšŒ ꡬ쑰와 κΆŒμœ„ 계측화 μΆœν˜„ 연ꡬ

semantic_scholarresearch

Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated Memory

μ œν•œλœ λ©”λͺ¨λ¦¬ μ˜ˆμ‚° λ‚΄μ—μ„œ κ²½ν—˜ 좕적을 ν†΅ν•œ μ˜¨λ””λ°”μ΄μŠ€ μ—μ΄μ „νŠΈ ν•™μŠ΅

semantic_scholarresearch

Universal Guideline-Driven Image Clustering via a Hybrid LLM Agent

ν…μŠ€νŠΈ μ§€μΉ¨μœΌλ‘œ λ‹€μ–‘ν•œ 이미지 ν΄λŸ¬μŠ€ν„°λ§ μž‘μ—…μ„ μˆ˜ν–‰ν•˜λŠ” 톡합 μ—μ΄μ „νŠΈ ν”„λ ˆμž„μ›Œν¬

semantic_scholarresearch

Explainable Control Framework (XCF) based on Fuzzy Model-Agnostic Explanation and LLM Agent-Supported Interface

λ³΅μž‘ν•œ μ œμ–΄ μ‹œμŠ€ν…œμ˜ λ™μž‘μ„ LLM μ—μ΄μ „νŠΈλ‘œ 인간이 이해할 수 μžˆλ„λ‘ μ„€λͺ