Learning GenAI via SOTA Papers - Explainer

Title: TokenMizer: Graph-Structured Session Memory for Long-Horizon LLM Context ManagementSource: http://arxiv.org/abs/2606.06337v1 Summary: TokenMizer introduces a graph-structured architectural primitive for managing long-horizon session memory, replacing inefficient flat-text history with a typed knowledge graph. This system achieves significant token compression while preserving the structural...

Title: The End of Software Engineering: How AI Agents Are Fundamentally Restructuring the Software Paradigm Source: http://arxiv.org/abs/2606.05608v1 Summary: This paper formalizes the shift from code-centric logic to LLM-driven reasoning loops, defining the emergent discipline of "Agentic Engineering." It provides a theoretical framework for self-evolving agent ecosystems and a roadmap...

Title: Scaling Self-Evolving Agents via Parametric Memory Source: http://arxiv.org/abs/2606.04536v1 Summary: This paper introduces a foundational framework for self-evolving agents that moves beyond static prompts by using online LoRA updates to adapt the model's parametric memory within a single episode. It establishes a new architectural paradigm where agents can genuinely learn and evolve t...

Title: Agentic Monte Carlo: Simulating Reinforcement Learning for Black-Box Agents Source: http://arxiv.org/abs/2606.05296v1 Summary: This work presents a foundational breakthrough for optimizing black-box LLM agents by applying the theoretical equivalence between reinforcement learning and Bayesian inference through Sequential Monte Carlo sampling. It enables principled, RL-style performance impr...

Title: Inducing Reasoning Primitives from Agent Traces Source: http://arxiv.org/abs/2606.02994v1 Summary: This work introduces a foundational reasoning loop that autonomously mines successful agent traces to codify recurrent reasoning moves into a reusable library of typed primitives. This method enables agents to systematically improve their own performance by discovering and abstracting their re...

Title: Agent libOS: A Library-OS-Inspired Runtime for Long-Running, Capability-Controlled LLM Agents Source: http://arxiv.org/abs/2606.03895v1 Summary: This paper introduces a novel architectural substrate that treats LLM agents as managed operating system processes with persistent identity, state, and capability-controlled resource access. It establishes a foundational runtime for long-running ag...

Title: OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web Agents Source: http://arxiv.org/abs/2606.02031v1 Summary: This paper introduces a comprehensive open framework for training visual web agents using online multi-turn reinforcement learning, overcoming the scalability limits of static datasets. It establishes a new state-of-the-art for open-source agents by provi...

Title: COMAP: Co-Evolving World Models and Agent Policies for LLM Agents Source: http://arxiv.org/abs/2606.02372v1 Summary: COMAP proposes a novel architectural primitive where textual world models and agent policies co-evolve through closed-loop interaction and self-distillation. This framework enables agents to adapt to dynamic environments by predicting future states and reflecting on action re...

Title: Adaptive Latent Agentic Reasoning Source: http://arxiv.org/abs/2606.02871v1 Summary: This paper introduces a dual-mode reasoning framework that dynamically alternates between compact latent reasoning and explicit chain-of-thought, optimizing the accuracy-efficiency trade-off for multi-turn agents. It establishes a significant architectural primitive by enabling agents to reserve heavy delib...

Title: SkillSmith: Co-Evolving Skills and Tools for Self-Improving Agent Systems Source: http://arxiv.org/abs/2606.01314v1 Summary: SkillSmith presents a foundational co-evolution framework that allows agents to simultaneously evolve their skill libraries and underlying toolsets through a synergy-aware reflection process. By utilizing an ecological utility model to manage skill-tool interactions,...

Title: Science Earth: Towards A Planet-Scale Operating System for AI-Native Scientific Discovery Source: http://arxiv.org/abs/2606.01316v1 Summary: It introduces a planet-scale scientific runtime and the EACN protocol, enabling decentralized AI capabilities to autonomously discover and negotiate task ownership. This shifts agentic orchestration from pre-designed workflows to emergent, self-correct...

Title: RLVR without Ineffective Samples: Group Prioritized Off-Policy Optimization for LLM Reasoning Source: http://arxiv.org/abs/2606.01281v1 Summary: This paper introduces POPO, a novel optimization framework that solves the critical zero-variance reward bottleneck in Reinforcement Learning with Verifiable Rewards (RLVR) for LLM reasoning. By implementing prioritized group replay and decoupled o...

Title: Learning to Adapt: Self-Improving Web Agent via Cognitive-Aware Exploration Source: http://arxiv.org/abs/2605.31365v1 Summary: SCALE introduces a foundational self-improving framework that enables agents to autonomously expand their cognitive boundaries through adversarial exploration and global planning strategies. It marks a significant shift from static, handcrafted execution pipelines t...

Title: EchoRL: Reinforcement Learning via Rollout Echoing Source: http://arxiv.org/abs/2605.31228v1 Summary: This paper introduces EchoRL, a novel reinforcement learning primitive that prevents training signal collapse in reasoning models by recovering gradients from successfully verified rollouts. It establishes a foundational method for post-training LLMs to achieve higher reasoning performance...

Title: GrepSeek: Training Search Agents for Direct Corpus InteractionSource: http://arxiv.org/abs/2605.29307v1 Summary: This paper introduces Direct Corpus Interaction (DCI), a foundational paradigm shift where search agents treat text corpora as executable environments via shell commands instead of traditional ranked indices. By training agents to find and compose evidence directly from raw data...

Title: ESPO: Early-Stopping Proximal Policy Optimization Source: http://arxiv.org/abs/2605.29860v1 Summary: Early-Stopping Proximal Policy Optimization (ESPO) provides a significant breakthrough in efficiency and reasoning for LLM reinforcement learning by detecting and terminating failed reasoning trajectories on-the-fly. This foundational optimization reduces compute overhead by 20% while improv...

Title: TRACER: Turn-level Regret Matching with Inner Reinforcement Credit for Cooperative Multi-LLM Reasoning Source: http://arxiv.org/abs/2605.28699v1 Summary: TRACER introduces a novel turn-level reinforcement framework that unifies regret matching with role-specific rewards to optimize multi-agent cooperation and reasoning. By separating the decision of when to speak from the content of the utt...

Title: Do Agents Think Deeper? A Mechanistic Investigation of Layer-Wise Dynamics in Sequential Planning Source: http://arxiv.org/abs/2605.27935v1 Summary: This study provides foundational mechanistic evidence that agentic reasoning requires dynamic, adaptive recruitment of model depth, distinguishing it from static inference tasks. These insights into layer-wise dynamics are critical for developi...

Title: Confidence-Orchestrated Self-Evolution against Uncertain LLM Feedback Source: http://arxiv.org/abs/2605.28010v1 Summary: COSE provides a foundational framework for LLM self-evolution by using intrinsic model confidence as an uncertainty signal to filter and weigh self-generated training signals. This approach addresses the critical bottleneck of error propagation in autonomous learning loop...

Title: MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation Source: http://arxiv.org/abs/2605.27366v1 Summary: This paper proposes a novel architectural framework for self-evolving agents that can autonomously create, store, and refine a library of reusable skills through a unified lifecycle management system. It introduces the concept of skill-level memory a...

Title: Innovation: An Almost Characterization of Hallucination Source: http://arxiv.org/abs/2605.26808v1 Summary: This work establishes a foundational probabilistic framework that formalizes hallucination as "innovation," providing a mathematical characterization of why LLMs produce outputs outside their training data. By deriving new lower bounds on hallucination rates based on "mi...

Title: Multi-Agent Coordination Adaptation via Structure-Guided Orchestration Source: http://arxiv.org/abs/2605.25746v1 Summary: The paper introduces a novel probabilistic framework for multi-agent coordination by casting orchestration as posterior inference over task-specific structures. This foundational approach balances structural stability with dynamic adaptability, significantly improving ex...

Title: How Should LLMs Consume High-Quality Data? Optimal Data Scheduling via Quality-Aware Functional Scaling Laws Source: http://arxiv.org/abs/2605.25698v1 Summary: This paper establishes foundational quality-aware functional scaling laws that provide the first theoretical closed-form solution for scheduling high-quality data during LLM training. The introduced 'Drop-Stable-Rampup' sched...

Title: Efficient Agentic Reasoning Through Self-Regulated Simulative Planning Source: http://arxiv.org/abs/2605.22138v1 Summary: This paper introduces a foundational three-system reasoning framework—comprising reactive, simulative, and self-regulated components—that enables agents to autonomously manage their planning depth and horizon. By treating the LLM as a world model for future-state predict...

Title: Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost Source: http://arxiv.org/abs/2605.22502v1 Summary: This work proposes the 'subterranean agent' paradigm, which replaces external orchestration frameworks by compiling agentic workflows directly into the model's weights via fine-tuning. This foundational shift addresses the co...