x

75% of people globally want unbiased news (Pew Research) but are concerned they're not getting that.

Get balanced coverage with Symby. See how stories are covered across diverse outlets and spot when one side covers a story the other skips.

PODCAST

The AI Research Deep Dive

The AI Research Deep Dive

From arXiv to insight: a daily tour of cutting-edge AI papers. The AI Research Deep Dive podcast dives into a new groundbreaking research paper every day. It combs through the most important details and results to give you a great idea of what the paper accomplishes and how it gets there.

science

All Episodes

00:16:19

DinoV3

The AI Research Deep Dive ·

2025/08/19

Arxiv: https://arxiv.org/abs/2508.10104v1This episode of "The AI Research Deep Dive" unpacks DINOv3, a state-of-the-art, self-supervised vision model from Meta AI. The host explains the fascinating problem the researchers faced when scaling up their models: as the model got better at understanding the big picture, its ability to perceive fine-grained details actually got worse. Listeners will learn about the paper's brilliant and intuitive solution, a new technique called "Gram

en

00:15:27

Sample More to Think Less: Group Filtered Policy...

The AI Research Deep Dive ·

2025/08/14

Arxiv: https://arxiv.org/abs/2508.09726This episode of "The AI Research Deep Dive" explores a Microsoft paper with a brilliant solution to a common AI problem: long, rambling, and repetitive answers. The paper, "Sample More to Think Less," introduces Group Filtered Policy Optimization (GFPO), a clever and counter-intuitive method to make AI models more concise. The host explains that to teach a model to "think less" and get to the point, GFPO makes it "sample

en

00:16:11

DataRater: Meta-Learned Dataset Curation

The AI Research Deep Dive ·

2025/08/12

This episode of "The AI Research Deep Dive" explores Google DeepMind's "DataRater," a paper that aims to turn the "black art" of data curation for LLMs into a data-driven science. The host explains how DataRater uses a clever meta-learning process to train a separate, smaller model whose only job is to rate the value of training data. Listeners will learn how this system moves beyond handwritten rules by learning to identify high-quality data that accelerates model

en

00:15:38

Scaling Up RL: Unlocking Diverse Reasoning in LLMs...

The AI Research Deep Dive ·

2025/08/07

Arxiv: https://arxiv.org/abs/2507.12507This episode of "The AI Research Deep Dive" unpacks an NVIDIA paper that offers a practical recipe for overcoming the common problem of "training plateaus" in reinforcement learning. The host breaks down how the researchers took a small 1.5-billion-parameter model and, through prolonged and stable training, made it competitive with specialized models in complex domains like math and coding. Listeners will learn about the core method, which

en

00:16:31

RLVMR: Reinforcement Learning with Verifiable...

The AI Research Deep Dive ·

2025/08/05

Arxiv: https://arxiv.org/abs/2507.22844This episode of "The AI Research Deep Dive" explores "RLVMR," a paper from Tencent that proposes a new way to build more reliable and intelligent AI agents. The host explains that instead of just rewarding an agent for completing a task, RLVMR rewards the agent for its reasoning process, teaching it to "think about its own thinking." Listeners will learn how the system uses a special vocabulary of "meta-reasoning"

en

00:16:49

X-Omni: Reinforcement Learning Makes Discrete...

The AI Research Deep Dive ·

2025/07/31

Arxiv: https://arxiv.org/abs/2507.22058This episode of "The AI Research Deep Dive" unpacks "X-Omni," a paper from Tencent that makes a bold claim: reinforcement learning can make autoregressive image models "great again." The host explains how this method tackles the historical weaknesses of autoregressive models, like blurry images and notoriously bad spelling. Listeners will learn about X-Omni's clever three-part architecture, which uses a large language model as

en

00:15:48

Groupe Sequence Policy Optimization

The AI Research Deep Dive ·

2025/07/29

Arxiv: https://www.arxiv.org/abs/2507.18071This episode of "The AI Research Deep Dive" unpacks "Group Sequence Policy Optimization" (GSPO), a new and powerful reinforcement learning algorithm from the creators of the Qwen models. The host explains how GSPO solves the critical problem of training instability and "model collapse" that plagues large-scale AI development. Listeners will learn the paper's core insight: a subtle but fundamental flaw in previous methods

en

00:17:03

Inverse Scaling in Test-Time Compute

The AI Research Deep Dive ·

2025/07/24

Arxiv: https://arxiv.org/abs/2507.14417This week on The AI Research Deep Dive, we unpack a provocative paper from Anthropic that asks: can an AI think too much? We've always assumed that longer reasoning leads to smarter answers, but new research shows this isn't always true. Forcing today's most advanced models to "think harder" can make them less accurate and amplify concerning behaviors, like expressing a desire for self-preservation. We explore the clever "trap"

en

00:19:29

DAPO: An Open-Source LLM Reinforcement Learning...

The AI Research Deep Dive ·

2025/07/22

Arxiv: https://arxiv.org/abs/2503.14476This episode of "The AI Research Deep Dive" unpacks the groundbreaking paper "DAPO: An Open-Source LLM Reinforcement Learning System at Scale," a significant release that democratizes state-of-the-art AI reasoning. The host explains how DAPO provides a fully open-source system that not only replicates but surpasses the performance of closed-door models on complex tasks like the AIME 2024 math competition, achieving this with greater

en

00:17:43

Proximal Policy Optimization

The AI Research Deep Dive ·

2025/07/17

Arxiv: https://arxiv.org/abs/1707.06347This podcast episode from "The A.I. Research Deep Dive" explores the landmark paper "Proximal Policy Optimization Algorithms," which introduced the robust and widely-used P.P.O. algorithm. The host explains how P.P.O. brilliantly solved the long-standing trade-off between simple but unstable policy gradient methods and stable but complex algorithms like T.R.P.O. Listeners will learn the core mechanism behind P.P.O.'s success: a clever

en

00:18:57

Reinforcement Learning with Action Chunking

The AI Research Deep Dive ·

2025/07/15

Arxiv: https://arxiv.org/abs/2507.07969This episode of The AI Research Deep Dive unpacks "Reinforcement Learning with Action Chunking," a paper that tackles the challenge of teaching robots complex, long-horizon tasks with sparse rewards. The host explains the paper's elegantly simple solution: instead of deciding on a single action at every millisecond, the agent learns to choose an entire sequence or "chunk" of actions at once. Listeners will learn how this one idea

en

00:15:50

Wider or Deeper? Scaling LLM Inference-Time Compute...

The AI Research Deep Dive ·

2025/07/10

Arxiv: https://arxiv.org/abs/2503.04412This episode of The AI Research Deep Dive tackles a fundamental question in AI problem-solving: is it better to go "wide" by trying many different solutions, or "deep" by iteratively refining a single one? The host unpacks a paper from Sakana AI that presents an elegant solution called Adaptive Branching Monte Carlo Tree Search (ABMCTS), an algorithm that intelligently decides at every step whether to explore a new path or exploit a

en

00:15:15

Mercury: Ultra-Fast Language Models Based on Diffusion

The AI Research Deep Dive ·

2025/07/08

Arxiv: https://arxiv.org/abs/2506.17298This episode of The AI Research Deep Dive unpacks "Mercury," a groundbreaking paper from Inception Labs that could fundamentally change how language models are built. The host explains how the Mercury model abandons the standard, one-word-at-a-time (autoregressive) approach used by models like GPT and instead adopts a diffusion-based method, inspired by image generation, to create entire blocks of text in parallel. This architectural shift results in a

en

00:17:41

GRPO aka DeepSeekMath: Pushing the Limits of...

The AI Research Deep Dive ·

2025/07/03

Arxiv link: https://arxiv.org/abs/2402.03300In this episode of "The AI Research Deep Dive," the host breaks down "DeepSeekMath," a groundbreaking paper that challenges the dominance of massive, proprietary AI models in mathematical reasoning. The discussion centers on how a relatively small, 7-billion-parameter open-source model managed to outperform Google's 540B-parameter Minerva, a model 77 times its size. Listeners will learn about the meticulous three-stage recipe for this success: an

en

00:14:22

Reinforcement Learning Teachers of Test Time Scaling

The AI Research Deep Dive ·

2025/07/01

Arxiv: https://www.arxiv.org/abs/2506.08388This week on The AI Research Deep Dive, we explore a groundbreaking paper from Sakana AI that flips the script on how we build reasoning models. For years, the approach has been to use massive, power-hungry models to stumble upon correct answers through Reinforcement Learning—an incredibly inefficient process. But what if we've been thinking about it all wrong? Sakana AI introduces the "Reinforcement-Learned Teacher" (RLT), a smaller model

en

00:15:26

Evolutionary Policy Optimization

The AI Research Deep Dive ·

2025/06/26

https://arxiv.org/abs/2503.19037This podcast episode from "The AI Research Deep Dive" unpacks the paper "Evolutionary Policy Optimization" (E.P.O.), a novel method designed to overcome the scalability limitations of traditional reinforcement learning algorithms like P.P.O. The host explains that E.P.O. creates a powerful hybrid system by combining the stability and efficiency of policy gradient methods with the diversity and scalability of evolutionary algorithms. It achieves this by using a single

en

00:16:11

The Gemini 2.5 Tech Report

The AI Research Deep Dive ·

2025/06/24

In a deep dive into Google DeepMind's "Gemini 2.5" technical report, this podcast episode explores a significant advancement in AI that pushes beyond simple instruction-following towards capable, goal-oriented agents. The host breaks down the paper's core innovations into three pillars: a more stable and efficient Sparse Mixture-of-Experts (MoE) architecture, a learned "Thinking" mechanism that allows the model to perform internal computations before answering, and

en

37 results

Similar Podcasts

Classic Radio Science Fiction

solgoodmedia.com

Mission to Mars

Inception Point Ai

The 365 Days of Astronomy

365DaysOfAstronomy.org

Creek Devil

william Jevning

UFO Chronicles Podcast

Nik Hunter