Flash Forward is a show about possible (and not so possible) future scenarios. What would the warranty on a sex robot look like? How would diplomacy work if we couldn’t lie? Could there ever be a fecal transplant black market? (Complicated, it wouldn’t, and yes, respectively, in case you’re curious.) Hosted and produced by award winning science journalist Rose Eveleth, each episode combines audio drama and journalism to go deep on potential tomorrows, and uncovers what those futures might re ...
…
continue reading
Player FM - Internet Radio Done Right
Checked 1d ago
Added two years ago
Content provided by Igor Melnyk. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by Igor Melnyk or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.
Similar to Arxiv Papers
Come dive into one of the curiously delightful conversations overheard at National Geographic’s headquarters, as we follow explorers, photographers, and scientists to the edges of our big, weird, beautiful world. Hosted by Peter Gwin and Amy Briggs.
…
continue reading
DNA science. Artificial intelligence. Smartphones and 3D printers. Science and technology have transformed the world we live in. But how did we get here? It wasn’t by accident. Well, sometimes it was. It was also the result of hard work, teamwork, and competition. And incredibly surprising moments. Hosted by bestselling author Steven Johnson (“How We Got To Now”), American Innovations uses immersive scenes to tell the stories of the scientists, engineers, and ordinary people behind the great ...
…
continue reading
Interviews with Anthropologists about their New Books Support our show by becoming a premium member! https://newbooksnetwork.supportingcast.fm/anthropology
…
continue reading
We humans could have a bright future ahead of us that lasts billions of years. But we have to survive the next 200 years first. Join Josh Clark of Stuff You Should Know for a 10-episode deep dive that explores the future of humanity and finds dangers we have never encountered before lurking just ahead. And if we humans are alone in the universe, if we don't survive intelligent life dies out with us too.
…
continue reading
Science Friction's latest series is: Brain Rot. We're looking at what being chronically online is doing to our brains. What's really going on with our attention spans and tech addiction? Is data-dumping your entire life into ChatGPT helpful? Can going internet free help you escape the doomscroll? And what's it like to be in love ... with an AI? National technology reporter Ange Lavoipierre tackles the wildest ways people are using tech and the big questions about our own use. That's Brain Ro ...
…
continue reading
The Skeptics' Guide to the Universe is a weekly science podcast discussing the latest science news, critical thinking, bad science, conspiracies and controversies. -The Skeptics' Guide to the Universe: Your escape to reality - Produced by SGU Productions, LLC: https://www.theskepticsguide.org
…
continue reading
Welcome to Curiosity Weekly from Discovery, hosted by Dr. Samantha Yammine. Once a week, we’ll bring you the latest and greatest in scientific discoveries and break down the details so that you don’t need a PhD to understand it. From neuroscience to climate tech to AI and genetics, no subject is off-limits. Join Sam as she interviews expert guests and investigates the research guiding some of the most exciting scientific breakthroughs affecting our world today. Hosted on Acast. See acast.com ...
…
continue reading
Join Dr Karl Kruszelnicki, Lucy Smith and their scientific guests, with a bunch of curious triple j listeners for a weekly injection of science, myth-bashing and answers!
…
continue reading
I discuss a variety of topics in both the natural and social sciences, exploring the many fascinating insights that the scientific method yields about the world around us.
…
continue reading
Player FM - Podcast App
Go offline with the Player FM app!
Go offline with the Player FM app!
))
Daily Deals
Podcasts Worth a Listen
SPONSORED
T
Tinfoil Swans
1 Encore: Will Poulter, Dave Beran, and The Bear 52:22
52:22 
Play Later 
Play Later 
Lists 
Like 
Liked52:22
Play Later
Play Later
Lists
Like
LikedSeason 3 of the smash hit FX/Hulu show “The Bear” roared to life just days ago, but Will Poulter (the actor who plays fan-favorite Luca) and 2014 F&W Best New Chef Dave Beran had been prepping for weeks. Poulter — like his co-star Jeremy Allen White — staged with Beran at his Santa Monica restaurant Pasjoli to learn how to accurately portray a professional chef onscreen. The lessons went so well, Beran says he’d hire Poulter as a cook — even despite a messy mishap with a pastry bag. The two dished all about getting kitchen culture right on and offscreen, what it takes to be at the top of your craft, and the pure magic of a great restaurant service. Learn more about your ad choices. Visit podcastchoices.com/adchoices…
[QA] OLMOTRACE: Tracing Language Model Outputs Back to Trillions of Training Tokens
Manage episode 476158571 series 3524393
Content provided by Igor Melnyk. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by Igor Melnyk or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.
OLMOTRACE is a real-time system that traces language model outputs to their training data, enabling users to explore fact-checking, hallucination, and creativity in language models.
https://arxiv.org/abs//2504.07096
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
2341 episodes
ShareManage episode 476158571 series 3524393
Content provided by Igor Melnyk. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by Igor Melnyk or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.
OLMOTRACE is a real-time system that traces language model outputs to their training data, enabling users to explore fact-checking, hallucination, and creativity in language models.
https://arxiv.org/abs//2504.07096
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
2341 episodes
All episodes
×
1 [QA] Sampling from Your Language Model One Byte at a Time 7:05
7:05 Play Later Play Later Lists Like Liked7:05
Play Later Play Later Lists Like LikedThis paper presents a method to convert autoregressive language models with BPE tokenizers into character-level models, addressing tokenization issues and enabling model interoperability and improved performance through ensemble and proxy-tuning. https://arxiv.org/abs//2506.14123 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 Sampling from Your Language Model One Byte at a Time 13:35
13:35 Play Later Play Later Lists Like Liked13:35
Play Later Play Later Lists Like LikedThis paper presents a method to convert autoregressive language models with BPE tokenizers into character-level models, addressing tokenization issues and enabling model interoperability and improved performance through ensemble and proxy-tuning. https://arxiv.org/abs//2506.14123 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] Don't throw the baby out with the bathwater: How and why deep learning for ARC 7:44
7:44 Play Later Play Later Lists Like Liked7:44
Play Later Play Later Lists Like LikedThis paper demonstrates that deep learning, through on-the-fly training and innovative techniques, significantly enhances performance on the Abstraction and Reasoning Corpus, achieving state-of-the-art results. https://arxiv.org/abs//2506.14276 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 Don't throw the baby out with the bathwater: How and why deep learning for ARC 32:30
32:30 Play Later Play Later Lists Like Liked32:30
Play Later Play Later Lists Like LikedThis paper demonstrates that deep learning, through on-the-fly training and innovative techniques, significantly enhances performance on the Abstraction and Reasoning Corpus, achieving state-of-the-art results. https://arxiv.org/abs//2506.14276 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] What Happens During the Loss Plateau? Understanding Abrupt Learning in Transformers 7:18
7:18 Play Later Play Later Lists Like Liked7:18
Play Later Play Later Lists Like LikedThis study explores abrupt learning in shallow Transformers, revealing a performance plateau characterized by repetition bias and representation collapse, with attention map learning as a critical bottleneck. https://arxiv.org/abs//2506.13688 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 What Happens During the Loss Plateau? Understanding Abrupt Learning in Transformers 19:43
19:43 Play Later Play Later Lists Like Liked19:43
Play Later Play Later Lists Like LikedThis study explores abrupt learning in shallow Transformers, revealing a performance plateau characterized by repetition bias and representation collapse, with attention map learning as a critical bottleneck. https://arxiv.org/abs//2506.13688 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention 8:28
8:28 Play Later Play Later Lists Like Liked8:28
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.13585 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
1 MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention 25:05
25:05 Play Later Play Later Lists Like Liked25:05
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.13585 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
1 [QA] Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation 8:10
8:10 Play Later Play Later Lists Like Liked8:10
Play Later Play Later Lists Like LikedWe present a diffusion-based framework for aligned novel view image and geometry generation, utilizing warping, inpainting, and cross-modal attention distillation for enhanced synthesis and prediction accuracy. https://arxiv.org/abs//2506.11924 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation 16:59
16:59 Play Later Play Later Lists Like Liked16:59
Play Later Play Later Lists Like LikedWe present a diffusion-based framework for aligned novel view image and geometry generation, utilizing warping, inpainting, and cross-modal attention distillation for enhanced synthesis and prediction accuracy. https://arxiv.org/abs//2506.11924 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] TreeRL: LLM Reinforcement Learning with On-Policy Tree Search 7:17
7:17 Play Later Play Later Lists Like Liked7:17
Play Later Play Later Lists Like LikedTreeRL is a novel reinforcement learning framework that integrates on-policy tree search, improving exploration and efficiency in reasoning tasks, outperforming traditional methods in benchmarks. https://arxiv.org/abs//2506.11902 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 TreeRL: LLM Reinforcement Learning with On-Policy Tree Search 19:00
19:00 Play Later Play Later Lists Like Liked19:00
Play Later Play Later Lists Like LikedTreeRL is a novel reinforcement learning framework that integrates on-policy tree search, improving exploration and efficiency in reasoning tasks, outperforming traditional methods in benchmarks. https://arxiv.org/abs//2506.11902 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] Solving Inequality Proofs with Large Language Models 8:20
8:20 Play Later Play Later Lists Like Liked8:20
Play Later Play Later Lists Like LikedThe paper addresses challenges in inequality proving for LLMs, introducing the INEQMATH dataset and a novel evaluation framework, revealing significant gaps in reasoning accuracy among leading models. https://arxiv.org/abs//2506.07927 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 Solving Inequality Proofs with Large Language Models 23:49
23:49 Play Later Play Later Lists Like Liked23:49
Play Later Play Later Lists Like LikedThe paper addresses challenges in inequality proving for LLMs, introducing the INEQMATH dataset and a novel evaluation framework, revealing significant gaps in reasoning accuracy among leading models. https://arxiv.org/abs//2506.07927 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
1 [QA] Reinforcement Learning Teachers of Test Time Scaling 7:54
7:54 Play Later Play Later Lists Like Liked7:54
Play Later Play Later Lists Like LikedThe paper introduces Reinforcement-Learned Teachers (RLTs) that enhance distillation efficiency by providing detailed explanations, outperforming larger models in reasoning tasks without requiring extensive exploration. https://arxiv.org/abs//2506.08388 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 Reinforcement Learning Teachers of Test Time Scaling 22:37
22:37 Play Later Play Later Lists Like Liked22:37
Play Later Play Later Lists Like LikedThe paper introduces Reinforcement-Learned Teachers (RLTs) that enhance distillation efficiency by providing detailed explanations, outperforming larger models in reasoning tasks without requiring extensive exploration. https://arxiv.org/abs//2506.08388 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers 7:05
7:05 Play Later Play Later Lists Like Liked7:05
Play Later Play Later Lists Like LikedThis study explores out-of-context reasoning in large language models, linking generalization and hallucination to a single mechanism, and formalizes it as a synthetic factual recall task. https://arxiv.org/abs//2506.10887 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers 18:28
18:28 Play Later Play Later Lists Like Liked18:28
Play Later Play Later Lists Like LikedThis study explores out-of-context reasoning in large language models, linking generalization and hallucination to a single mechanism, and formalizes it as a synthetic factual recall task. https://arxiv.org/abs//2506.10887 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Spurious Rewards: Rethinking Training Signals in RLVR 7:41
7:41 Play Later Play Later Lists Like Liked7:41
Play Later Play Later Lists Like LikedReinforcement learning with verifiable rewards (RLVR) enhances mathematical reasoning in Qwen2.5-Math, achieving notable performance improvements, but spurious rewards may not benefit other models like Llama3 or OLMo2. https://arxiv.org/abs//2506.10947 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 Spurious Rewards: Rethinking Training Signals in RLVR 30:13
30:13 Play Later Play Later Lists Like Liked30:13
Play Later Play Later Lists Like LikedReinforcement learning with verifiable rewards (RLVR) enhances mathematical reasoning in Qwen2.5-Math, achieving notable performance improvements, but spurious rewards may not benefit other models like Llama3 or OLMo2. https://arxiv.org/abs//2506.10947 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generation 8:08
8:08 Play Later Play Later Lists Like Liked8:08
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.09991 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
A
Arxiv Papers
1 Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generation 24:30
24:30 Play Later Play Later Lists Like Liked24:30
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.09991 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
Reinforcement Pre-Training (RPT) enhances language models by using reinforcement learning for next-token prediction, improving accuracy and providing a strong foundation for further fine-tuning. https://arxiv.org/abs//2506.08007 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
Reinforcement Pre-Training (RPT) enhances language models by using reinforcement learning for next-token prediction, improving accuracy and providing a strong foundation for further fine-tuning. https://arxiv.org/abs//2506.08007 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
https://arxiv.org/abs//2506.06215 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
A
Arxiv Papers
https://arxiv.org/abs//2506.06215 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
A
Arxiv Papers
The paper presents UNDO, a method that enhances unlearning in LLMs through distillation, achieving robust capability removal with reduced compute and data requirements compared to traditional retraining methods. https://arxiv.org/abs//2506.06278 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
The paper presents UNDO, a method that enhances unlearning in LLMs through distillation, achieving robust capability removal with reduced compute and data requirements compared to traditional retraining methods. https://arxiv.org/abs//2506.06278 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
This paper introduces log-linear attention, enhancing linear attention's efficiency by using a logarithmically growing set of hidden states, improving sequence modeling while maintaining computational efficiency. https://arxiv.org/abs//2506.04761 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
This paper introduces log-linear attention, enhancing linear attention's efficiency by using a logarithmically growing set of hidden states, improving sequence modeling while maintaining computational efficiency. https://arxiv.org/abs//2506.04761 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening 7:45
7:45 Play Later Play Later Lists Like Liked7:45
Play Later Play Later Lists Like LikedThis paper critiques GRPO's bias in training language models for theorem proving and introduces the unlikeliness reward to enhance performance and sample diversity, achieving competitive results. https://arxiv.org/abs//2506.02355 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening 16:56
16:56 Play Later Play Later Lists Like Liked16:56
Play Later Play Later Lists Like LikedThis paper critiques GRPO's bias in training language models for theorem proving and introduces the unlikeliness reward to enhance performance and sample diversity, achieving competitive results. https://arxiv.org/abs//2506.02355 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
The Self-Challenging framework enables agents to generate and train on high-quality tasks autonomously, achieving significant performance improvements using self-generated data in tool-use benchmarks. https://arxiv.org/abs//2506.01716 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
The Self-Challenging framework enables agents to generate and train on high-quality tasks autonomously, achieving significant performance improvements using self-generated data in tool-use benchmarks. https://arxiv.org/abs//2506.01716 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Why Gradients Rapidly Increase Near the End of Training 7:00
7:00 Play Later Play Later Lists Like Liked7:00
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.02285 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
A
Arxiv Papers
1 Why Gradients Rapidly Increase Near the End of Training 11:24
11:24 Play Later Play Later Lists Like Liked11:24
Play Later Play Later Lists Like Likedhttps://arxiv.org/abs//2506.02285 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
A
Arxiv Papers
1 [QA] GEM: Empowering LLM for both Embedding Generation and Language Understanding 7:41
7:41 Play Later Play Later Lists Like Liked7:41
Play Later Play Later Lists Like LikedThe paper introduces GEM, a self-supervised method enabling decoder-only LLMs to generate high-quality text embeddings, enhancing performance on embedding benchmarks while preserving original text generation capabilities. https://arxiv.org/abs//2506.04344 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 GEM: Empowering LLM for both Embedding Generation and Language Understanding 20:38
20:38 Play Later Play Later Lists Like Liked20:38
Play Later Play Later Lists Like LikedThe paper introduces GEM, a self-supervised method enabling decoder-only LLMs to generate high-quality text embeddings, enhancing performance on embedding benchmarks while preserving original text generation capabilities. https://arxiv.org/abs//2506.04344 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] HYPERSTEER: Activation Steering at Scale with Hypernetworks 7:49
7:49 Play Later Play Later Lists Like Liked7:49
Play Later Play Later Lists Like LikedHYPERSTEER introduces hypernetwork architectures for generating effective steering vectors in language models, outperforming existing methods and achieving strong performance on unseen prompts. Code available at GitHub. https://arxiv.org/abs//2506.03292 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 HYPERSTEER: Activation Steering at Scale with Hypernetworks 9:15
9:15 Play Later Play Later Lists Like Liked9:15
Play Later Play Later Lists Like LikedHYPERSTEER introduces hypernetwork architectures for generating effective steering vectors in language models, outperforming existing methods and achieving strong performance on unseen prompts. Code available at GitHub. https://arxiv.org/abs//2506.03292 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
The OpenThoughts project creates open-source datasets for reasoning models, achieving state-of-the-art results with OpenThinker3-7B, trained on 1.2M examples, available at openthoughts.ai. https://arxiv.org/abs//2506.04178 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
The OpenThoughts project creates open-source datasets for reasoning models, achieving state-of-the-art results with OpenThinker3-7B, trained on 1.2M examples, available at openthoughts.ai. https://arxiv.org/abs//2506.04178 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Accelerating Diffusion LLMs via Adaptive Parallel Decoding 8:08
8:08 Play Later Play Later Lists Like Liked8:08
Play Later Play Later Lists Like LikedThe paper introduces adaptive parallel decoding (APD), enhancing diffusion large language models' speed by dynamically adjusting token sampling, improving throughput while maintaining quality compared to autoregressive models. https://arxiv.org/abs//2506.00413 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 Accelerating Diffusion LLMs via Adaptive Parallel Decoding 21:09
21:09 Play Later Play Later Lists Like Liked21:09
Play Later Play Later Lists Like LikedThe paper introduces adaptive parallel decoding (APD), enhancing diffusion large language models' speed by dynamically adjusting token sampling, improving throughput while maintaining quality compared to autoregressive models. https://arxiv.org/abs//2506.00413 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
A
Arxiv Papers
1 [QA] Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning 7:34
7:34 Play Later Play Later Lists Like Liked7:34
Play Later Play Later Lists Like LikedThis paper presents a self-reflection and reinforcement learning method that enhances large language models' performance on complex tasks, achieving significant improvements even with limited feedback. https://arxiv.org/abs//2505.24726 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers…
Welcome to Player FM!
Player FM is scanning the web for high-quality podcasts for you to enjoy right now. It's the best podcast app and works on Android, iPhone, and the web. Signup to sync subscriptions across devices.
Daily Deals
Daily Deals
Similar to Arxiv Papers
Flash Forward is a show about possible (and not so possible) future scenarios. What would the warranty on a sex robot look like? How would diplomacy work if we couldn’t lie? Could there ever be a fecal transplant black market? (Complicated, it wouldn’t, and yes, respectively, in case you’re curious.) Hosted and produced by award winning science journalist Rose Eveleth, each episode combines audio drama and journalism to go deep on potential tomorrows, and uncovers what those futures might re ...
…
continue reading
Come dive into one of the curiously delightful conversations overheard at National Geographic’s headquarters, as we follow explorers, photographers, and scientists to the edges of our big, weird, beautiful world. Hosted by Peter Gwin and Amy Briggs.
…
continue reading
DNA science. Artificial intelligence. Smartphones and 3D printers. Science and technology have transformed the world we live in. But how did we get here? It wasn’t by accident. Well, sometimes it was. It was also the result of hard work, teamwork, and competition. And incredibly surprising moments. Hosted by bestselling author Steven Johnson (“How We Got To Now”), American Innovations uses immersive scenes to tell the stories of the scientists, engineers, and ordinary people behind the great ...
…
continue reading
Interviews with Anthropologists about their New Books Support our show by becoming a premium member! https://newbooksnetwork.supportingcast.fm/anthropology
…
continue reading
We humans could have a bright future ahead of us that lasts billions of years. But we have to survive the next 200 years first. Join Josh Clark of Stuff You Should Know for a 10-episode deep dive that explores the future of humanity and finds dangers we have never encountered before lurking just ahead. And if we humans are alone in the universe, if we don't survive intelligent life dies out with us too.
…
continue reading
Science Friction's latest series is: Brain Rot. We're looking at what being chronically online is doing to our brains. What's really going on with our attention spans and tech addiction? Is data-dumping your entire life into ChatGPT helpful? Can going internet free help you escape the doomscroll? And what's it like to be in love ... with an AI? National technology reporter Ange Lavoipierre tackles the wildest ways people are using tech and the big questions about our own use. That's Brain Ro ...
…
continue reading
The Skeptics' Guide to the Universe is a weekly science podcast discussing the latest science news, critical thinking, bad science, conspiracies and controversies. -The Skeptics' Guide to the Universe: Your escape to reality - Produced by SGU Productions, LLC: https://www.theskepticsguide.org
…
continue reading
Welcome to Curiosity Weekly from Discovery, hosted by Dr. Samantha Yammine. Once a week, we’ll bring you the latest and greatest in scientific discoveries and break down the details so that you don’t need a PhD to understand it. From neuroscience to climate tech to AI and genetics, no subject is off-limits. Join Sam as she interviews expert guests and investigates the research guiding some of the most exciting scientific breakthroughs affecting our world today. Hosted on Acast. See acast.com ...
…
continue reading
Join Dr Karl Kruszelnicki, Lucy Smith and their scientific guests, with a bunch of curious triple j listeners for a weekly injection of science, myth-bashing and answers!
…
continue reading
I discuss a variety of topics in both the natural and social sciences, exploring the many fascinating insights that the scientific method yields about the world around us.
…
continue reading
Player FM - Podcast App
Go offline with the Player FM app!
Go offline with the Player FM app!
