Igor Melnyk
Arxiv Papers
Running out of time to catch up with new arXiv papers? We take the most impactful papers and present them as convenient podcasts. If you're a visual learner, we offer these papers in an engaging video format. Our service fills the gap between overly brief paper summaries and time-consuming full paper reads. You gain academic insights in a time-efficient, digestible format. Code behind this work: https://github.com/imelnyk/ArxivPapers
Não deixe de visitar o site do podcast e apoiar quem o produz: github.com
Onde ouvir?
Podcasts no app Replaio Radio Em breveOs podcasts estão chegando ao app. Instale agora e seja o primeiro a descobrir um jeito totalmente novo de curtir podcasts
Instale grátis
Android 5 mi+ downloads · nota 4,8
iOS em breve
Episódios
[short] Evolutionary Optimization of Model Merging Recipes 23.03.2024 1:54
Evolutionary algorithms automate creating powerful foundation models by merging diverse open-source models, achieving state-of-the-art performance in Japanese language and culture tasks. https://arxiv.org/abs//2403.13187 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:...
Evolutionary Optimization of Model Merging Recipes 23.03.2024 23:10
Evolutionary algorithms automate creating powerful foundation models by merging diverse open-source models, achieving state-of-the-art performance in Japanese language and culture tasks. https://arxiv.org/abs//2403.13187 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:...
[short] Recourse for Reclamation: Chatting with Generative Language Models 23.03.2024 1:59
Toxicity scoring in language models can hinder information access and cultural norms. A novel recourse mechanism allows users to set toxicity thresholds dynamically for improved interaction. https://arxiv.org/abs//2403.14467 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spot...
Recourse for Reclamation: Chatting with Generative Language Models 23.03.2024 8:32
Toxicity scoring in language models can hinder information access and cultural norms. A novel recourse mechanism allows users to set toxicity thresholds dynamically for improved interaction. https://arxiv.org/abs//2403.14467 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Spot...
[short] A Unified Framework for Model Editing 22.03.2024 1:49
The paper introduces a framework unifying ROME and MEMIT model editing techniques under a preservation-memorization objective, presenting EMMET as a new batched memory-editing algorithm for Transformers. https://arxiv.org/abs//2403.14236 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id16...
A Unified Framework for Model Editing 22.03.2024 13:54
The paper introduces a framework unifying ROME and MEMIT model editing techniques under a preservation-memorization objective, presenting EMMET as a new batched memory-editing algorithm for Transformers. https://arxiv.org/abs//2403.14236 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id16...
[short] Language Models Can Reduce Asymmetry in Information Markets 22.03.2024 1:47
The paper explores the buyer's inspection paradox in information markets using simulated agents with language models, focusing on biases, pricing effects, and quality outcomes. https://arxiv.org/abs//2403.14443 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...
Language Models Can Reduce Asymmetry in Information Markets 22.03.2024 14:34
The paper explores the buyer's inspection paradox in information markets using simulated agents with language models, focusing on biases, pricing effects, and quality outcomes. https://arxiv.org/abs//2403.14443 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...
[short] AI and Memory Wall 22.03.2024 2:42
Unsupervised training data and neural scaling have led to larger models, but memory bandwidth is now the main bottleneck in AI applications, requiring architecture and strategy redesign. https://arxiv.org/abs//2403.14123 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:...
AI and Memory Wall 22.03.2024 9:52
Unsupervised training data and neural scaling have led to larger models, but memory bandwidth is now the main bottleneck in AI applications, requiring architecture and strategy redesign. https://arxiv.org/abs//2403.14123 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:...
RewardBench: Evaluating Reward Models for Language Modeling 21.03.2024 19:52
The paper introduces REWARDBENCH, a dataset and code-base for evaluating reward models in aligning language models with human preferences, shedding light on their capabilities and limitations. https://arxiv.org/abs//2403.13787 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016 Sp...
Evaluating Frontier Models for Dangerous Capabilities 21.03.2024 23:54
https://arxiv.org/abs//2403.13793 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
Reverse Training to Nurse the Reversal Curse 21.03.2024 13:28
Large language models struggle with generalizing from "A has B" to "B is a feature of A" due to Zipf's law. Reverse training doubles tokens, improving performance on reversal tasks. https://arxiv.org/abs//2403.13799 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id...
[short] Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs 20.03.2024 1:50
Technique proposed to transfer reasoning capabilities from large-language models to vision-language models, achieving state-of-the-art performance on ChartQA, PlotQA, and FigureQA tasks. https://arxiv.org/abs//2403.12596 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:...
Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs 20.03.2024 17:55
Technique proposed to transfer reasoning capabilities from large-language models to vision-language models, achieving state-of-the-art performance on ChartQA, PlotQA, and FigureQA tasks. https://arxiv.org/abs//2403.12596 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:...
[short] Vid2Robot: End-to-end Video-conditioned Policy Learning with Cross-Attention Transformers 20.03.2024 2:03
Vid2Robot enables robots to learn tasks from human video demonstrations, outperforming other methods by 20% and showcasing potential for real-world applications. https://arxiv.org/abs//2403.12943 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.spoti...
Vid2Robot: End-to-end Video-conditioned Policy Learning with Cross-Attention Transformers 20.03.2024 17:35
Vid2Robot enables robots to learn tasks from human video demonstrations, outperforming other methods by 20% and showcasing potential for real-world applications. https://arxiv.org/abs//2403.12943 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.spoti...
[short] Yell At Your Robot Improving On-the-Fly from Language Corrections 20.03.2024 2:11
Hierarchical policies combining language and low-level control improve robotic task performance. Human feedback on language corrections enhances high-level policies, leading to significant performance gains in dexterous manipulation tasks. https://arxiv.org/abs//2403.12910 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.ap...
Yell At Your Robot Improving On-the-Fly from Language Corrections 20.03.2024 18:53
Hierarchical policies combining language and low-level control improve robotic task performance. Human feedback on language corrections enhances high-level policies, leading to significant performance gains in dexterous manipulation tasks. https://arxiv.org/abs//2403.12910 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.ap...
[short] Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillation 19.03.2024 1:42
Latent Adversarial Diffusion Distillation (LADD) improves image synthesis speed and quality compared to existing methods by leveraging generative features from pretrained latent diffusion models. https://arxiv.org/abs//2403.12015 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016...
Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillation 19.03.2024 13:38
Latent Adversarial Diffusion Distillation (LADD) improves image synthesis speed and quality compared to existing methods by leveraging generative features from pretrained latent diffusion models. https://arxiv.org/abs//2403.12015 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016...
[short] Larimar: Large Language Models with Episodic Memory Control 19.03.2024 2:19
Larimar introduces a brain-inspired architecture for efficient knowledge updating in Large Language Models, achieving accuracy comparable to baselines with 4-10x speed-ups. https://arxiv.org/abs//2403.11901 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://podca...
Larimar: Large Language Models with Episodic Memory Control 19.03.2024 20:45
Larimar introduces a brain-inspired architecture for efficient knowledge updating in Large Language Models, achieving accuracy comparable to baselines with 4-10x speed-ups. https://arxiv.org/abs//2403.11901 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://podca...
[short] Mind Eye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data 19.03.2024 2:25
Novel method enables high-quality visual perception reconstructions from minimal fMRI data by pretraining across subjects and using functional alignment, improving generalization and achieving state-of-the-art results. https://arxiv.org/abs//2403.11207 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/ar...
Mind Eye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data 19.03.2024 16:14
Novel method enables high-quality visual perception reconstructions from minimal fMRI data by pretraining across subjects and using functional alignment, improving generalization and achieving state-of-the-art results. https://arxiv.org/abs//2403.11207 YouTube: https://www.youtube.com/@ArxivPapers TikTok: https://www.tiktok.com/@arxiv_papers Apple Podcasts: https://podcasts.apple.com/us/podcast/ar...
Podcasts parecidos
Ouça o podcast Arxiv Papers no Replaio
Rádio e podcasts em um só app - grátis e sem cadastro. Instale hoje e não perca o lançamento
O Replaio não é o publicador dos podcasts; os nomes dos programas, as capas e o áudio pertencem aos seus autores e são distribuídos por feeds RSS públicos
