Read/listen to the newsletter/podcast on substack:
My Twitter, LinkedIn, Discord, Patreon, consultation booking page, etc:
Timestamps:
00:00 intro
00:58 Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers
01:48 COLLAGE – Collaborative Human-Agent Interaction Generation using Hierarchical Latent Diffusion and LMs
02:15 The Perfect Blend – Redefining RLHF with Mixture of Judges
04:38 A Looming Replication Crisis in Evaluating Behavior in LMs? Evidence and Solutions
05:18 1 Trillion Token (1TT) Platform – A Novel Framework for Efficient Data Sharing and Compensation in LLMs
06:27 Counter-Current Learning – A Biologically Plausible Dual Network Approach for Deep Learning
07:49 Unifying back-propagation and forward-forward algorithms through model predictive control
09:08 Can LLMs Really Learn to Translate a Low-Resource Language from One Grammar Book?
09:45 HM3 – Hierarchical Multi-Objective Model Merging for Pretrained Models
10:22 Hierarchical Federated ADMM
10:51 Cottention – Linear Transformers With Cosine Attention
15:18 Effects of AI Feedback on Learning, the Skill Gap, and Intellectual Diversity
18:57 Kolmogorov-Arnold Network AEs
20:12 PeerArg – Argumentative Peer Review with LLMs
21:55 When a LM is optimized for reasoning, does it still show embers of autoregression – An analysis of OpenAI o1
25:12 Adaptive Inference-Time Compute – LLMs Can Predict if They Can Do Better, Even Mid-Generation
26:29 LLMs Know More Than They Show – On the Intrinsic Representation of LLM Hallucinations
27:06 Selective Attention Improves Transformer
27:59 On the Proper Treatment of Tokenization in Psycholinguistics
28:32 FAN – Fourier Analysis Networks
30:25 Generalization emerges from local optimization in a self-organized learning network
32:35 Fair Decentralized Learning
33:18 Intelligence at the Edge of Chaos
35:11 Post-edits Are Preferences Too
36:06 Theoretical Insights into Fine-Tuning Attention – Generalization and Optimization
37:12 EmbedLLM – Learning Compact Representations of LLMs
37:53 Planning in Strawberry Fields – Evaluating and Improving the Planning and Scheduling Capabilities of LRM o1
38:33 Mitigating Memorization In LMs
39:22 U-shaped and Inverted-U Scaling behind Emergent Abilities of LLMs
40:36 ENTP – Encoder-only Next Token Prediction
42:07 House of Cards – Massive Weights in LLMs
44:16 Geometric Signatures of Compositionality Across a LM’s Lifetime
45:13 FlashMask – Efficient and Rich Mask Extension of FlashAttention
46:06 Sparse AEs Reveal Temporal Difference Learning in LLMs
46:37 nGPT – Normalized Transformer with Representation Learning on the Hypersphere
50:15 Draft on the Fly – Adaptive Self-Speculative Decoding using Cosine Similarity
50:49 Investigating the Synergistic Effects of Dropout and Residual Connections on LM Training
51:43 Do Music Generation Models Encode Music Theory?
52:08 RisingBALLER – A path towards a foundational model for football players data analytics
53:24 Self-Updatable LLMs with Parameter Integration
53:50 Stability analysis of chaotic systems in latent spaces
54:30 MoS – Unleashing Parameter Efficiency of LoRA with Mixture of Shards
55:51 Are LLMs Aware that Some Questions are not Open-ended?
56:56 TikGuard – A Deep Learning Transformer-Based Solution for Detecting Unsuitable TikTok Content for Kids
59:48 Vision LMs See What You Want but not What You See
[ad_2]
source
-
-
-
-
