[ Ссылка ]

Feel free to read all the summaries over on my substack weekly newsletter:
[ Ссылка ]

To see/use my workflow for automating these weekly paper summaries click here:
[ Ссылка ]

—————————————
Join my Discord Server to talk with other Tunadorks about this nerdy stuff
[ Ссылка ]

Read my weekly newsletter on the newest AI papers here:
[ Ссылка ]

Timestamps:
0:00 Intro
0:38 What Are We Optimizing For? A Human-centric Evaluation Of Deep Learning-based Recommender Systems
1:41 Visibility into AI Agents
2:02 Universal Neurons in GPT2 Language Models
4:02 UniMS-RAG - A Unified Multi-source Retrieval-Augmented Generation for Personalized Dialogue Systems
5:24 Understanding the Generalization Benefits of Late Learning Rate Decay
5:50 Understanding Video Transformers via Universal Concept Discovery
6:50 Transformers and Cortical Waves - Encoders for Pulling In Context Across Time
7:57 Transformer-Based Models Are Not Yet Perfect At Learning to Emulate Structural Recursion
9:24 Towards a Systems Theory of Algorithms
9:52 Towards Uncertainty-Aware Language Agent
10:44 Topologies of Reasoning - Demystifying Chains, Trees, and Graphs of Thoughts
12:32 The twin peaks of learning neural networks
13:02 The Definitive Guide to Policy Gradients in Deep Reinforcement Learning - Theory, Algorithms and Implementations
13:40 Temporal Blind Spots in Large Language Models
14:57 ServerlessLLM - Locality-Enhanced Serverless Inference for Large Language Models
15:18 Reinforcement learning for question answering in programming domain using public community scoring as a human feedback
16:45 Reframing Offline Reinforcement Learning as a Regression Problem
17:18 Over-Reasoning and Redundant Calculation of Large Language Models
18:13 Optimisation in Neurosymbolic Learning Systems
19:02 New Pathways in Coevolutionary Computation
19:33 Multimodal Pathway - Improve Transformers with Irrelevant Data from Other Modalities
20:56 MambaByte - Token-free Selective State Space Model
21:28 MM-LLMs - Recent Advances in MultiModal Large Language Models
23:40 Large Language Models are Superpositions of All Characters - Attaining Arbitrary Role-play via Self-Alignment
24:33 Language Models as Hierarchy Encoders
25:20 In-context Learning with Retrieved Demonstrations for Language Models - A Survey
25:54 How AI Ideas Affect the Creativity, Diversity, and Evolution of Human Ideas - Evidence From a Large, Dynamic Experiment
28:37 Hierarchical Prompts for Rehearsal-free Continual Learning
29:05 Hallucination is Inevitable - An Innate Limitation of Large Language Models
29:58 Graph Condensation - A Survey
30:21 Gaussian Adaptive Attention is All You Need - Robust Contextual Representations Across Multiple Modalities
31:36 From Understanding to Utilization - A Survey on Explainability for Large Language Models
32:10 FLLIC - Functionally Lossless Image Compression
33:09 Exploring the Truth and Beauty of Theory Landscapes with Machine Learning
33:26 Dynamic Long-Term Time-Series Forecasting via Meta Transformer Networks
33:56 Critical Data Size of Language Models from a Grokking Perspective
35:33 Comparing Human-Centered Language Modeling - Is it Better to Model Groups, Individual Traits, or Both?
36:27 Can A Cognitive Architecture Fundamentally Enhance LLMs? Or Vice Versa?
37:08 Bridging Evolutionary Algorithms and Reinforcement Learning - A Comprehensive Survey
37:24 Backpropagation Through Agents
38:17 Automated facial recognition system using deep learning for pain assessment in adults with cerebral palsy
38:32 Assessing and Understanding Creativity in Large Language Models
39:41 Anisotropy Is Inherent to Self-Attention in Transformers
40:36 Active Inference as a Model of Agency
41:14 ALMs - Authorial Language Models for Authorship Attribution
41:38 AI for social science and social science of AI - A Survey
42:33 A survey on recent advances in named entity recognition
42:50 A Survey of Deep Learning and Foundation Models for Time Series Forecasting
43:02 A Reply to Makelov et al. (2023)'s "Interpretability Illusion" Arguments
43:39 "All of Me" - Mining Users' Attributes from their Public Spotify Playlists

свернуть