A survey that unifies prior work on multi-agent LLM systems via the LIFE framework, mapping dependencies across collaboration, failure attribution, and autonomous self-evolution while identifying cross-stage challenges.
hub Canonical reference
Advances and challenges in foundation agents: From brain-inspired intelligence to evolutionary, collaborative, and safe systems
Canonical reference. 86% of citing Pith papers cite this work as background.
29
Pith papers citing it
Background
86% of classified citations
abstract
The advent of large language models (LLMs) has catalyzed a transformative shift in artificial intelligence, paving the way for advanced intelligent agents capable of sophisticated reasoning, robust perception, and versatile action across diverse domains. As these agents increasingly drive AI research and practical applications, their design, evaluation, and continuous improvement present intricate, multifaceted challenges. This book provides a comprehensive overview, framing intelligent agents within modular, brain-inspired architectures that integrate principles from cognitive science, neuroscience, and computational research. We structure our exploration into four interconnected parts. First, we systematically investigate the modular foundation of intelligent agents, systematically mapping their cognitive, perceptual, and operational modules onto analogous human brain functionalities and elucidating core components such as memory, world modeling, reward processing, goal, and emotion. Second, we discuss self-enhancement and adaptive evolution mechanisms, exploring how agents autonomously refine their capabilities, adapt to dynamic environments, and achieve continual learning through automated optimization paradigms. Third, we examine multi-agent systems, investigating the collective intelligence emerging from agent interactions, cooperation, and societal structures. Finally, we address the critical imperative of building safe and beneficial AI systems, emphasizing intrinsic and extrinsic security threats, ethical alignment, robustness, and practical mitigation strategies necessary for trustworthy real-world deployment. By synthesizing modular AI architectures with insights from different disciplines, this survey identifies key research challenges and opportunities, encouraging innovations that harmonize technological advancement with meaningful societal benefit.
hub tools
citation-role summary
background 12
dataset 2
citation-polarity summary
representative citing papers
VideoDR is a new benchmark for open-web video deep research that tests multimodal models on cross-frame visual anchor extraction, interactive retrieval, and multi-hop reasoning over joint video-web evidence.
FOREVER aligns replay intervals in LLM continual learning with a model-centric time based on optimizer update magnitudes and an Ebbinghaus-inspired forgetting curve to reduce catastrophic forgetting.
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
AgentForesight introduces an online auditor model that predicts decisive errors in multi-agent trajectories at the earliest step using a coarse-to-fine reinforcement learning recipe on a new curated dataset AFTraj-2K.
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
ReCast repairs all-zero groups and uses contrastive updates on strongest positives and hardest negatives to improve RL in generative recommendation, yielding up to 36.6% better Pass@1 with only 4.1% of baseline rollout budget.
Mem-π is a framework using a dedicated model and decision-content decoupled RL to generate context-specific guidance on demand for LLM agents, outperforming retrieval baselines by over 30% on web navigation.
ReBel uses belief-consistency supervision and belief-aware grouping to improve credit assignment in long-horizon RL for LLM agents, achieving up to 20.4 percentage points higher success and 2.1x better sample efficiency than GRPO on ALFWorld and WebShop.
LPDS quantifies difficulty of logic-preserving problem variations and searches for the hardest ones, producing up to 5x larger performance drops than random sampling and better robustness gains from fine-tuning on difficult examples.
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
FuzzAgent deploys specialized agents that collaborate on harness generation, execution, and crash triage to evolve fuzzing campaigns, delivering 45-191% more branch coverage than four baselines on 20 C/C++ libraries and surfacing 102 real bugs.
CHAL is a multi-agent dialectic system that performs structured belief optimization over defeasible domains using Bayesian-inspired graph representations and configurable meta-cognitive value system hyperparameters.
DiffMAS jointly optimizes latent communication and reasoning in multi-agent LLM systems via parameter-efficient supervised training on trajectories, yielding consistent gains over baselines on math, science, and code benchmarks.
A new benchmark shows LLM smartphone agents achieve comparable success with screen text alone as with screenshots, but both fail often due to UI accessibility and reasoning gaps.
Agent-GWO uses collaborative grey-wolf-inspired agents to jointly optimize LLM prompts and decoding settings, yielding higher accuracy and stability than prior single-agent prompt optimization methods on math and hybrid reasoning benchmarks.
ADAM extracts data from LLM agent memory with up to 100% attack success rate by estimating data distribution and selecting queries via entropy guidance.
The paper maps agent memory research via three forms (token-level, parametric, latent), three functions (factual, experiential, working), and dynamics of formation/evolution/retrieval, plus benchmarks and future directions.
Generalizable agents require environment scaling via diverse executable rule-sets, distinguished from trajectory and task scaling in a new taxonomy.
SynthAgent uses dual refinement of synthetic tasks and trajectories to produce higher-quality training data that improves web agent adaptation to target environments.
Mujica-MyGo decomposes multi-turn RAG interactions via multi-agent workflows and applies minimalist policy gradient optimization to improve performance on QA benchmarks while avoiding long-context problems.
UI-TARS-2 reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld while attaining 59.8 mean normalized score on a 15-game suite through multi-turn RL and scalable data generation.
A comprehensive review of self-evolving AI agents that improve themselves over time, organized via a framework of inputs, agent system, environment, and optimizers, with domain-specific and safety discussions.
A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.
citing papers explorer
-
Beyond Individual Intelligence: Surveying Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems
A survey that unifies prior work on multi-agent LLM systems via the LIFE framework, mapping dependencies across collaboration, failure attribution, and autonomous self-evolution while identifying cross-stage challenges.
-
Watching, Reasoning, and Searching: A Video Deep Research Benchmark on Open Web for Agentic Video Reasoning
VideoDR is a new benchmark for open-web video deep research that tests multimodal models on cross-frame visual anchor extraction, interactive retrieval, and multi-hop reasoning over joint video-web evidence.
-
FOREVER: Forgetting Curve-Inspired Memory Replay for Language Model Continual Learning
FOREVER aligns replay intervals in LLM continual learning with a model-centric time based on optimizer update magnitudes and an Ebbinghaus-inspired forgetting curve to reduce catastrophic forgetting.
-
Harnessing Agentic Evolution
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
-
AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems
AgentForesight introduces an online auditor model that predicts decisive errors in multi-agent trajectories at the earliest step using a coarse-to-fine reinforcement learning recipe on a new curated dataset AFTraj-2K.
-
Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
-
ReCast: Recasting Learning Signals for Reinforcement Learning in Generative Recommendation
ReCast repairs all-zero groups and uses contrastive updates on strongest positives and hardest negatives to improve RL in generative recommendation, yielding up to 36.6% better Pass@1 with only 4.1% of baseline rollout budget.
-
Mem-$\pi$: Adaptive Memory through Learning When and What to Generate
Mem-π is a framework using a dedicated model and decision-content decoupled RL to generate context-specific guidance on demand for LLM agents, outperforming retrieval baselines by over 30% on web navigation.
-
Rewarding Beliefs, Not Actions: Consistency-Guided Credit Assignment for Long-Horizon Agents
ReBel uses belief-consistency supervision and belief-aware grouping to improve credit assignment in long-horizon RL for LLM agents, achieving up to 20.4 percentage points higher success and 2.1x better sample efficiency than GRPO on ALFWorld and WebShop.
-
LPDS: Evaluating LLM Robustness Through Logic-Preserving Difficulty Scaling
LPDS quantifies difficulty of logic-preserving problem variations and searches for the hardest ones, producing up to 5x larger performance drops than random sampling and better robustness gains from fine-tuning on difficult examples.
-
The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
-
FuzzAgent: Multi-Agent System for Evolutionary Library Fuzzing
FuzzAgent deploys specialized agents that collaborate on harness generation, execution, and crash triage to evolve fuzzing campaigns, delivering 45-191% more branch coverage than four baselines on 20 C/C++ libraries and surfacing 102 real bugs.
-
CHAL: Council of Hierarchical Agentic Language
CHAL is a multi-agent dialectic system that performs structured belief optimization over defeasible domains using Bayesian-inspired graph representations and configurable meta-cognitive value system hyperparameters.
-
Learning to Communicate: Toward End-to-End Optimization of Multi-Agent Language Systems
DiffMAS jointly optimizes latent communication and reasoning in multi-agent LLM systems via parameter-efficient supervised training on trajectories, yielding consistent gains over baselines on math, science, and code benchmarks.
-
Do LLMs Need to See Everything? A Benchmark and Study of Failures in LLM-driven Smartphone Automation using Screentext vs. Screenshots
A new benchmark shows LLM smartphone agents achieve comparable success with screen text alone as with screenshots, but both fail often due to UI accessibility and reasoning gaps.
-
Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language Models
Agent-GWO uses collaborative grey-wolf-inspired agents to jointly optimize LLM prompts and decoding settings, yielding higher accuracy and stability than prior single-agent prompt optimization methods on math and hybrid reasoning benchmarks.
-
ADAM: A Systematic Data Extraction Attack on Agent Memory via Adaptive Querying
ADAM extracts data from LLM agent memory with up to 100% attack success rate by estimating data distribution and selecting queries via entropy guidance.
-
Memory in the Age of AI Agents
The paper maps agent memory research via three forms (token-level, parametric, latent), three functions (factual, experiential, working), and dynamics of formation/evolution/retrieval, plus benchmarks and future directions.
-
Scalable Environments Drive Generalizable Agents
Generalizable agents require environment scaling via diverse executable rule-sets, distinguished from trajectory and task scaling in a new taxonomy.
-
SynthAgent: Adapting Web Agents with Synthetic Supervision
SynthAgent uses dual refinement of synthetic tasks and trajectories to produce higher-quality training data that improves web agent adaptation to target environments.
-
Advancing Multi-Agent RAG Systems with Minimalist Reinforcement Learning
Mujica-MyGo decomposes multi-turn RAG interactions via multi-agent workflows and applies minimalist policy gradient optimization to improve performance on QA benchmarks while avoiding long-context problems.
-
UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning
UI-TARS-2 reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld while attaining 59.8 mean normalized score on a 15-game suite through multi-turn RL and scalable data generation.
-
A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems
A comprehensive review of self-evolving AI agents that improve themselves over time, organized via a framework of inputs, agent system, environment, and optimizers, with domain-specific and safety discussions.
-
Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models
A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.
-
Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges
A survey that taxonomizes threats to agentic AI, reviews benchmarks and evaluation methods, discusses technical and governance defenses, and identifies open challenges.
-
Multi-Agent Systems: From Classical Paradigms to Large Foundation Model-Enabled Futures
A survey comparing classical multi-agent systems with large foundation model-enabled multi-agent systems, showing how the latter enables semantic-level collaboration and greater adaptability.
-
Bridging Brains and Machines: A Unified Frontier in Neuroscience, Artificial Intelligence, and Neuromorphic Systems
A position and survey paper that identifies convergence between neuroscience, AGI, and neuromorphic computing and outlines four key integration challenges.
- MemORAI: Memory Organization and Retrieval via Adaptive Graph Intelligence for LLM Conversational Agents
- Perspective on Bias in Biomedical AI: Preventing Downstream Healthcare Disparities