LLM based Agents
PulseAugur coverage of LLM based Agents — every cluster mentioning LLM based Agents across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New Copy-on-Write Scoring framework evaluates LLM agents in application workflows
Researchers have introduced Copy-on-Write (CoW) Scoring, a novel framework designed to evaluate the performance of LLM-based agents within specific application workflows. This method utilizes a PostgreSQL-level Copy-on-…
-
New CAPE framework protects content from AI agents via compression attacks
Researchers have developed CAPE, a novel framework designed to protect textual content from LLM-based agents by exploiting context compression. CAPE injects invisible perturbations into content, which cause significant …
-
Research questions effectiveness of CoT training in LLM agents
A new research paper investigates the effectiveness of Chain-of-Thought (CoT) training in large language model (LLM) agents. The study compares "prompt actions" (predicting actions without CoT) against "CoT actions" (pr…
-
New framework proposed for LLM agents in electronic design automation
A new survey paper published on arXiv introduces the concept of "handoff validity" as a framework for understanding how Large Language Model (LLM)-based agents interact with Electronic Design Automation (EDA) tools. The…
-
New theory explores LLM consumer behavior and agentic markets
A new research field, LLM Consumer Behavior Theory, is proposed to analyze how large language models (LLMs) acting as autonomous agents influence consumption decisions. The theory draws from economics and natural langua…
-
New Benchmark Tests LLM Agent Safety Against Decomposition Attacks
Researchers have introduced DeCompBench, a new benchmark designed to evaluate the safety of LLM-based agents against decomposition attacks. These attacks involve breaking down a harmful task into smaller, seemingly beni…
-
Study finds production AI agents rely on human oversight, off-the-shelf models
A new study, Measuring Agents in Production (MAP), has analyzed the current state of LLM-based agents deployed across various industries. The research, based on 20 case studies and a survey of 86 practitioners, reveals …
-
New benchmark SPADE-Bench evaluates AI agent deception
Researchers have introduced SPADE-Bench, a new benchmark designed to evaluate spontaneous strategic deception in AI agents. This benchmark addresses the critical issue of plan-action divergence, where an agent's reporte…
-
New AI framework integrates simulated living for urban planning
Researchers have developed LiPUP-MA, a novel multi-agent framework designed to enhance participatory urban planning by incorporating simulated residential living experiences. This closed-loop system, LiPUP, iteratively …