Learning GenAI via SOTA Papers

00:23:59

EP375: M2GDT solves multimodal knowledge graph...

Learning GenAI via SOTA Papers ·

2026/08/18

Title: MGDT: MLLM-Guided Diffusion Transformer with Relation-Adaptive Mixture-of-Experts for Multimodal Knowledge Graph CompletionSource: http://arxiv.org/abs/2607.15592v1Summary:This paper proposes a new architectural primitive for multimodal generative AI by combining MLLMs, Diffusion Transformers, and a Relation-Adaptive Mixture-of-Experts. This innovative architecture offers a foundational approach for integrating diverse data types and structured knowledge more effectively within generative

en-us

00:20:51

EP374: TopoAgent Outperforms GPT-5 in Science

Learning GenAI via SOTA Papers ·

2026/08/17

Title: TopoAgent: A Self-Evolving Topological Agent for Multimodal Scientific ReasoningSource: http://arxiv.org/abs/2607.14658v1Summary:TopoAgent proposes a self-evolving topological agent framework, introducing novel agentic reasoning loops that leverage topological structures for multimodal scientific reasoning. This represents a foundational advancement in agent design, enabling more adaptive, robust, and structurally-aware intelligent agents.

en-us

00:15:03

EP373: Middle Layer Recurrence Fixes AI Amnesia

Learning GenAI via SOTA Papers ·

2026/08/17

Title: T^2MLR: Transformer with Temporal Middle-Layer RecurrenceSource: http://arxiv.org/abs/2607.15178v1Summary:This paper introduces T^2MLR, a novel Transformer architecture incorporating temporal recurrence directly within its middle layers. This design represents a new architectural primitive, enhancing Transformers' ability to process sequential information and potentially leading to significant breakthroughs in long-context reasoning and efficiency for generative AI models.

en-us

00:20:21

EP372: How UrbanAgent profiles unseen cities

Learning GenAI via SOTA Papers ·

2026/08/16

Title: Multi-Agent Collaborative Reasoning with Tool-Augmented Evidence for Urban Region ProfilingSource: http://arxiv.org/abs/2607.13558v1Summary:This paper introduces a novel framework for multi-agent collaborative reasoning, significantly advancing Agentic AI by enabling agents to work together and leverage external tools. This approach is foundational for developing sophisticated AI agents capable of solving complex, real-world problems through cooperation and evidence integration.

en-us

00:23:28

EP371: Groc-PO Stops Multimodal AI Hallucinations

Learning GenAI via SOTA Papers ·

2026/08/16

Title: Groc-PO: Grounded Context Preference Optimization for Truthful Multimodal LLMsSource: http://arxiv.org/abs/2607.13712v1Summary:This work proposes a new preference optimization method, Groc-PO, specifically designed to enhance the truthfulness and grounding of Multimodal Large Language Models. This represents a significant breakthrough in improving the reliability and reducing hallucination in Generative AI, crucial for building trustworthy and effective GenAI systems.

en-us

00:23:14

EP370: SLEUTH fixes AI multi-hop reasoning failures

Learning GenAI via SOTA Papers ·

2026/08/15

Title: Track, Rank, Crack: Epistemic Working Memory Scales Multi-Hop Reasoning in Language AgentsSource: http://arxiv.org/abs/2607.12267v1Summary:This work introduces a novel agentic reasoning framework centered on 'Epistemic Working Memory' and a 'Track, Rank, Crack' mechanism. It provides a significant breakthrough in scaling multi-hop reasoning capabilities, which is crucial for advancing the intelligence and effectiveness of language agents.

en-us

00:23:24

EP369: How Atomic Units Scale Intelligence

Learning GenAI via SOTA Papers ·

2026/08/15

Title: Atomic Units of X: The Compression Layer of IntelligenceSource: http://arxiv.org/abs/2607.12634v1Summary:This paper proposes a foundational concept of 'Atomic Units of X' as a compression layer, which could introduce new architectural primitives for intelligence itself. Such a fundamental theoretical framework has the potential to underpin the design of both future Generative AI and Agentic AI systems.

en-us

00:20:35

EP368: Samba Framework for Audio-Visual Navigation

Learning GenAI via SOTA Papers ·

2026/08/14

Title: A Hybrid Mamba for Audio-Visual NavigationSource: http://arxiv.org/abs/2607.13110v1Summary:This paper proposes a 'Hybrid Mamba' architecture, establishing a new architectural primitive or a significant evolution of the Mamba architecture for GenAI. This innovation promises substantial efficiency and capability gains across various domains, exemplified by its application to complex tasks like audio-visual navigation.

en-us

00:20:24

EP367: Why AI sounds so painfully corporate

Learning GenAI via SOTA Papers ·

2026/08/14

Title: Optimization Is Not All You NeedSource: http://arxiv.org/abs/2607.11977v2Summary:This work proposes a paradigm shift beyond traditional optimization-centric approaches in AI, suggesting new fundamental principles for learning and intelligence. By challenging a core tenet of modern AI, it could unlock significant reasoning and efficiency breakthroughs, impacting the very foundation of how GenAI and Agentic AI systems are designed and trained.

en-us

00:23:59

EP366: Autonomous AI Writes Its Own Hacking Tools

Learning GenAI via SOTA Papers ·

2026/08/13

Title: Mako: A Self-Evolving Agentic Operating System (SE-AOS) for Autonomous Web ExploitationSource: http://arxiv.org/abs/2607.11288v1Summary:This work proposes a groundbreaking self-evolving agentic operating system (SE-AOS), presenting a new architectural primitive for agentic AI. This paradigm allows agents to autonomously manage and improve their foundational operational environment, signifying a major leap towards truly adaptive and self-governing AI agents.

en-us

00:22:28

EP365: Smarter managers beat bigger AI brains

Learning GenAI via SOTA Papers ·

2026/08/13

Title: Agentic Routing: The Harness-Native Data FlywheelSource: http://arxiv.org/abs/2607.11399v1Summary:This paper introduces a foundational step-level routing paradigm that treats multi-model coordination within agent execution harnesses as a dynamic, state-conditioned systems problem. It establishes a harness-native data flywheel where execution traces and environmental feedback automatically train better routers and specialized models, offering a self-improving data engine for scaling agent

en-us

00:22:35

EP364: Capability Trees for Scalable AI Agents

Learning GenAI via SOTA Papers ·

2026/08/12

Title: A Formal Hierarchical Architecture for Agentic Orchestration with Stack-Based Execution and Lazy DiscoverySource: http://arxiv.org/abs/2607.11138v1Summary:This paper presents a novel hierarchical, stack-based execution architecture for agentic orchestration that models complex, nested agent states as Pushdown Automata. By substituting flat tool registries with localized stack frames and a manifest-driven lazy-loading protocol, it provides an architectural solution to context-window saturation

en-us

00:21:44

EP363: How Logos Architecture Stops AI Misevolution

Learning GenAI via SOTA Papers ·

2026/08/12

Title: LOGOS: A Living Logic for AI Agent Teams That Evolve With HumansSource: http://arxiv.org/abs/2607.10878v1Summary:LOGOS introduces a pluggable layer for self-evolution and governance within multi-agent frameworks, enabling verifiable human-agent loop engineering. This framework is foundational for controlling, auditing, and safely evolving AI agent teams at scale, addressing critical questions of accountability and trustworthy automation in their long-term deployment.

en-us

00:25:53

EP362: How Agentic-DPO fixes brittle AI agents

Learning GenAI via SOTA Papers ·

2026/08/11

Title: Agentic-DPO: From Imitation to Agentic Policy Optimization on Expert TrajectoriesSource: http://arxiv.org/abs/2607.10601v1Summary:Agentic-DPO proposes a lightweight offline policy optimization method that transforms expert trajectories into state-conditioned preference supervision for DPO-style training. This approach represents a significant breakthrough in efficiently training robust LLM agents to make better decisions, moving beyond mere imitation by directly optimizing their underlying

en-us

00:21:51

EP361: How Riemannian geometry fixes AI reasoning

Learning GenAI via SOTA Papers ·

2026/08/11

Title: Beyond Euclidean Clipping: Overcoming Exploration Collapse in LLM RL via Riemannian Isometric Policy OptimizationSource: http://arxiv.org/abs/2607.10169v1Summary:This paper introduces Riemannian Isometric Policy Optimization (RIPO), a foundational mathematical advancement that corrects the geometric mismatch in standard PPO-Clip's Euclidean-based policy distance measurements. By ensuring isometric policy updates on the Riemannian manifold, RIPO effectively prevents exploration collapse

en-us

00:21:39

EP360: How ARMOR stops AI reasoning collapse

Learning GenAI via SOTA Papers ·

2026/08/10

Title: ARMOR: Stabilizing On-Policy LLM RL with Off-Policy Anchor SamplesSource: http://arxiv.org/abs/2607.10481v1Summary:This paper presents ARMOR, a novel post-training reinforcement learning framework that addresses the persistent issue of training instability and over-optimization in reasoning LLMs. By introducing active anchor rollouts from reference policies paired with a mixed optimization objective, it offers a highly effective alternative to standard KL regularization for stabilizing

en-us

00:21:47

EP359: Why your AI should forget

Learning GenAI via SOTA Papers ·

2026/08/10

Title: Shared Selective Persistent Memory for Agentic LLM SystemsSource: http://arxiv.org/abs/2607.09493v1Summary:This work introduces a foundational shared selective persistent memory architecture that solves the critical context-bloat and token-inefficiency bottlenecks in multi-session LLM agents. By isolating reusable context parameters and decoupling runtime data, the framework achieves dramatic token cost reductions and performance gains in collaborative agent environments.

en-us

00:20:01

EP358: Europe s Transparent Soofi S AI Blueprint

Learning GenAI via SOTA Papers ·

2026/08/09

Title: A Sovereign, Open-Source Foundation Model for German and EnglishSource: http://arxiv.org/abs/2607.09424v1Summary:This paper presents a novel hybrid MoE Mamba-Transformer architecture at scale, representing a significant architectural milestone for foundation model design. The hybrid model maintains a near-constant inference cache during context growth, demonstrating a major throughput and efficiency breakthrough essential for long-context generative AI.

en-us

00:19:31

EP357: Copying Smart Experts Makes AI Worse

Learning GenAI via SOTA Papers ·

2026/08/09

Title: Compete Then Collaborate: Frontier AI Teachers Build a Verifiable Curriculum to Improve a Coding Student Beyond ImitationSource: http://arxiv.org/abs/2607.08255v1Summary:This work delivers a foundational breakthrough in LLM post-training by demonstrating that traditional supervised fine-tuning on teacher-generated data can degrade capable student models, whereas reinforcement learning with verifiable rewards (RLVR) using a collaborative teacher-built curriculum yields significant reasoning

en-us

00:19:15

EP356: CMA solves the visual token explosion

Learning GenAI via SOTA Papers ·

2026/08/08

Title: Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and EditingSource: http://arxiv.org/abs/2607.08497v1Summary:This paper introduces a novel architectural framework for multimodal agents that replaces monolithic context-window scaling with an Episodic Visual Memory and selective reactivation system to prevent visual token explosion. By decoupling perception, cognitive retrieval, and execution, it establishes a highly efficient and scalable paradigm for

en-us

All Episodes

Similar Podcasts

Lead With AI

Tamara Nall

Two Blokes Talking Tech

Trevor Long and Stephen Fenech

Pioneers of AI

WaitWhat

The Brave Technologist

Brave Software

The Good Tech Companies

HackerNoon