EP224: HaM-World

EP229: Fixing AI Overthinking

Title: LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models Source: http://arxiv.org/abs/2605.09806v1 Summary:LEAD establishes a foundational reinforcement learning mechanism for reasoning models that dynamically calibrates the balance between correctness and verbosity at each training step. It solves the critical issue of 'overthinking' in modern reasoning models by introducing online, per-problem length estimation, paving the way for more efficient and scalable reasoning architectures.

EP228: Do Self-Evolving Agents Forget

Title: Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation Source: http://arxiv.org/abs/2605.09315v1 Summary: This paper introduces the 'capability erosion' framework to quantify how autonomous self-evolution can degrade an agent's prior knowledge across workflows and models. It proposes Capability-Preserving Evolution (CPE) as a necessary architectural constraint for building stable, lifelong learning agents that can adapt to new tasks without catastrophic forgetting.

EP227: FlowAgent Continuous Flow

Title: Tools as Continuous Flow for Evolving Agentic ReasoningSource: http://arxiv.org/abs/2605.07339v1 Summary: FlowAgent reconceptualizes agentic reasoning by replacing discrete, step-wise tool orchestration with continuous trajectory generation using conditional flow matching. This foundational framework provides theoretical guarantees for error attenuation and global planning, representing a significant shift in how agents execute long-horizon reasoning tasks.

EP226: Unlimited AI Thinking

Title: Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models Source: http://arxiv.org/abs/2605.07721v1 Summary:This paper introduces a novel architectural primitive that decouples reasoning depth from memory consumption in looped language models, enabling constant-memory iterative reasoning. By sharing a single KV cache across loops via a learnable gating mechanism, it provides a foundational efficiency breakthrough for models performing multi-step computation in embedding space.

EP225: The LOVER Framework

Title: Logic-Regularized Verifier Elicits Reasoning from LLMs Source: http://arxiv.org/abs/2605.05893v1 Summary: This work presents a novel reasoning framework that uses logical consistency rules to regularize unsupervised verifiers, eliminating the need for expensive supervised datasets. By treating verification as a binary latent variable problem, it achieves performance comparable to supervised models in eliciting complex reasoning from off-the-shelf LLMs.