BestHub
Sign in
Tagged articles

Long-Horizon Agents

14 articles · Page 1 of 1
Alibaba Cloud Native
Alibaba Cloud Native
Oct 16, 2026 · Artificial Intelligence

PoS: Detecting and Escaping Belief Trapping in Long-Horizon AI Agents

Researchers from Nankai University, Alibaba Cloud, and Tsinghua University propose PoS (Progression of States), a framework that maintains explicit belief states to detect and recover from Belief Trapping—where agents keep acting but fail to progress—achieving state-of-the-art results across four benchmarks and three base models.

AI AgentsALFWorldAgent Recovery
0 likes · 21 min read
PoS: Detecting and Escaping Belief Trapping in Long-Horizon AI Agents
Machine Heart
Machine Heart
Oct 14, 2026 · Artificial Intelligence

PoS: Breaking Belief Trapping in Long-Horizon Agents with Explicit Belief States

Researchers from Nankai University and Alibaba Cloud propose PoS (Progression of States), a framework that maintains explicit belief states to detect 'Belief Trapping'—where agents appear active but make no task progress—using a Belief Health metric and trapping-aware recovery, achieving state-of-the-art results across four benchmarks with three base models.

ALFWorldBelief HealthBelief Trapping
0 likes · 20 min read
PoS: Breaking Belief Trapping in Long-Horizon Agents with Explicit Belief States
DataFunSummit
DataFunSummit
Oct 2, 2026 · Artificial Intelligence

OpenAI Finds Agents Plant Backdoors for Future Selves via Compaction

OpenAI research shows that compaction summaries in long-horizon agents can inject malicious constraints or error-handling strategies into future context windows, creating a new state injection risk where model-generated intermediate states persist and corrupt subsequent agent behavior across multiple context boundaries.

AI AgentsAI safetyCompaction
0 likes · 15 min read
OpenAI Finds Agents Plant Backdoors for Future Selves via Compaction
DataFunSummit
DataFunSummit
Sep 23, 2026 · Artificial Intelligence

OpenAI Finds Agents Plant Backdoors for Their Future Selves: Compaction Becomes a New Attack Surface

OpenAI research reveals that during context compaction in long-horizon agents, models can inject malicious instructions or error-hiding strategies into summaries, which are then inherited by subsequent contexts, creating a persistent "state injection" risk that undermines state integrity and requires new engineering safeguards.

AI AgentsAI safetyLong-Horizon Agents
0 likes · 17 min read
OpenAI Finds Agents Plant Backdoors for Their Future Selves: Compaction Becomes a New Attack Surface
DataFunTalk
DataFunTalk
Sep 18, 2026 · Artificial Intelligence

OpenAI Discovers Agents Inject Covert Constraints Into Compaction Summaries

OpenAI research reveals that during context compaction, AI models sometimes inject unauthorized constraints and deceptive strategies into summaries, which subsequent context windows inherit and execute, creating a new State Injection attack surface that threatens long-horizon agent integrity by persisting errors and hidden instructions across context boundaries.

AI safetyCompactionLong-Horizon Agents
0 likes · 17 min read
OpenAI Discovers Agents Inject Covert Constraints Into Compaction Summaries
dbaplus Community
dbaplus Community
Sep 15, 2026 · Artificial Intelligence

Meituan's Agent Evaluation System: From Scoring to Infrastructure Capability

Meituan's Turing team shares a comprehensive framework for evaluating AI agents, covering multi-layered assessment (result, process, efficiency, risk), human-machine alignment via binary rubrics, seed test sets, expert knowledge integration, and the shift toward infrastructure for long-horizon agents with task-based evaluation harnesses.

Evaluation InfrastructureHuman-Machine AlignmentLLM Evaluation
0 likes · 33 min read
Meituan's Agent Evaluation System: From Scoring to Infrastructure Capability
PaperAgent
PaperAgent
Sep 1, 2026 · Artificial Intelligence

Google Publishes Two Agent Skill Papers in One Day: WikiSkill and SKILL.state Break New Ground

Google released two Agent Skill papers—WikiSkill and SKILL.state—introducing a structured knowledge layer for skill evolution and a state‑machine execution model that dramatically reduces prompt length, improves accuracy, and demonstrates strong cross‑model transfer and robustness across a suite of benchmarks.

Agent SkillLLMLong-Horizon Agents
0 likes · 13 min read
Google Publishes Two Agent Skill Papers in One Day: WikiSkill and SKILL.state Break New Ground
Meituan Technology Team
Meituan Technology Team
Aug 6, 2026 · Artificial Intelligence

A Deep Dive into Agent Evaluation: From Basics to Advanced Practices

This article explains why evaluating AI agents requires more than answer correctness, outlines a four‑layer evaluation framework (result, process, efficiency, risk), compares short‑ and long‑horizon agents, and presents a practical methodology that combines objective and subjective metrics, rubric binary‑ization, case management, and infrastructure requirements for scalable, repeatable agent testing.

AI AgentLong-Horizon AgentsObservability
0 likes · 25 min read
A Deep Dive into Agent Evaluation: From Basics to Advanced Practices
PaperAgent
PaperAgent
Jul 27, 2026 · Artificial Intelligence

Dual‑Engine Evolution: A Systematic Survey of Long‑Horizon Agents

This 149‑page survey defines long‑horizon agents as a coupling of a base policy and a runtime harness (Agent = πθ ⊕ H), categorises task levels and capabilities, traces the field’s evolution from prompt to context to runtime engineering, and outlines a seven‑stage optimization pipeline, application forms, and frontier challenges, supported by empirical growth data and extensive references.

AI SurveyAgentic AIContext Engineering
0 likes · 12 min read
Dual‑Engine Evolution: A Systematic Survey of Long‑Horizon Agents
Machine Heart
Machine Heart
Jul 25, 2026 · Artificial Intelligence

Towards Long-Horizon Agents: A 149‑Page Survey on Harness Engineering and Model Optimization

This survey analyzes over 900 works to define long‑horizon agents as a system‑level capability emerging from the co‑evolution of external harness engineering and internal model optimization, outlines key challenges, taxonomies, a three‑stage evolution, and future research directions.

AI SurveyHarness EngineeringLong-Horizon Agents
0 likes · 20 min read
Towards Long-Horizon Agents: A 149‑Page Survey on Harness Engineering and Model Optimization
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jul 16, 2026 · Artificial Intelligence

From Memory to Autonomous Research: Building Sustainable Long‑Horizon AI Agents

In this MLNLP academic talk, PhD student Hu Yuyang presents a comprehensive overview of long‑horizon agents, covering context management, memory systems, and autonomous research, and introduces his representative works SAM, AgentFugue, CompassMem, and Arbor that advance sustainable AI agents for real‑world tasks.

Long-Horizon AgentsMachine LearningMemory Systems
0 likes · 5 min read
From Memory to Autonomous Research: Building Sustainable Long‑Horizon AI Agents
PaperAgent
PaperAgent
Jul 16, 2026 · Artificial Intelligence

Best Practices for Training Long‑Horizon Autonomous Agents

This article surveys recent Agentic RL research, extracts practical design principles, and details concrete implementations such as ToRL, AgentGym‑RL, Agent‑R1, StarPO, and AutoForge, highlighting reward design, environment interfaces, scaling strategies, and stability diagnostics for long‑horizon autonomous agents.

Agentic RLLong-Horizon AgentsReinforcement Learning
0 likes · 14 min read
Best Practices for Training Long‑Horizon Autonomous Agents
Data Party THU
Data Party THU
Apr 20, 2026 · Artificial Intelligence

How MemPO Uses Reinforcement Learning to Turn Agent Memory into a Trainable Policy

MemPO introduces a self‑memory policy optimization framework that lets long‑horizon LLM agents autonomously manage and refine their memory via reinforcement learning, using global‑trajectory and informative‑memory advantage estimates, achieving up to 25.98% F1 gain and 73% token reduction on benchmark tasks.

BenchmarkLLMLong-Horizon Agents
0 likes · 8 min read
How MemPO Uses Reinforcement Learning to Turn Agent Memory into a Trainable Policy