برچسب: عامل‌های هوشمند

PerMIT: Protecting Persistent Memory for LLM Agents with Content Confirmation and Community Trust

PerMIT is a memory architecture for LLM agents that keeps unconfirmed environment-derived content behind a confirmation gate, machine-checks a specified single-user core in Lean 4, and proposes logged-confirmation-based…

PACT: Testing Enterprise AI Agent Compliance Under Pressure

PACT is a benchmark testing whether enterprise language-model agents and assistants follow workplace and regulatory rules under urgency, user pressure, and multi-turn pushback.

Hypothesis-Driven Autonomous Materials Synthesis with Multimodal LLM Agents

SynAgent, a multimodal AI-agent system, created analysis tools in a physical automated laboratory and used active hypothesis testing to identify a favorable growth window for LiCoO₂ films.

LiteReality-Agent: Agentic Reconstruction of Interactive 3D Indoor Scenes

LiteReality-Agent uses a coding agent to turn RGB-D room scans into editable, simulation-ready 3D scenes. Results reported on 10 scenes indicate improved visual similarity and geometric consistency, but…

Argo-Bench: Evaluating Data Agents in Enterprise Workflows

Argo-Bench is a simulated environment for evaluating data agents that must navigate a large ERP warehouse, perform analysis, and submit results as actions or reports.

Global Coherence in Multi-Agent Collaboration

This paper studies cases where each agent acts correctly from its local view, yet the team’s combined result violates shared constraints. Its central proposal is an external harness…

Mem++: Non-Destructive Long-Term Memory for Organizational Agents

Mem++ preserves complete organizational documents with dates and authors, deferring the choice among conflicting versions until question answering.

Mingbird: A Local-First Harness for Small-Model Agents

This preprint examines whether improving an agent harness, without changing the model, can make small open models more reliable on real tasks in Windows.