Parameterized Complexity of s-Club Cluster Edge Deletion
arxiv.org·16h
🕸️Network Topology
Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention
huggingface.co·12h·
Discuss: Hacker News
📊Learned Metrics
The key to conversational speech recognition
datasciencecentral.com·1h
🎵Audio ML
Efficient and accurate search in petabase-scale sequence repositories
nature.com·1d·
Discuss: Hacker News
🔄Burrows-Wheeler
Comparing human and language models sentence processing difficulties on complex structures
arxiv.org·16h
🔗Parser Combinators
Computational Complexity of Covering Two-vertex Multigraphs with Semi-edges
arxiv.org·16h
🕸️Graph Algorithms
Less Is More: Recursive Reasoning with Tiny Networks
github.com·1d·
Discuss: Hacker News
📊Quantization
The Hidden Oracle Inside Your AI: Unveiling Data Density with Latent Space Magic by Arvind Sundararajan
dev.to·18h·
Discuss: DEV
🧠Machine Learning
Nearly Instance-Optimal Parameter Recovery from Many Trajectories via Hellinger Localization
arxiv.org·16h
🧮Kolmogorov Bounds
Channel Simulation and Distributed Compression with Ensemble Rejection Sampling
arxiv.org·1d
Information Bottleneck
Partial Information Decomposition via Normalizing Flows in Latent Gaussian Distributions
arxiv.org·2d
🧠Machine Learning
Three Solutions to Nondeterminism in AI
blog.hellas.ai·1d·
Discuss: Hacker News
🎯Performance Proofs
Mathematical model of memory suggests 7 senses are optimal
skoltech.ru·3h·
Discuss: Hacker News
🔲Cellular Automata
Quantifying the Accuracy-Interpretability Trade-Off in Concept-Based Sidechannel Models
arxiv.org·1d
🧠Machine Learning
Mind the Gap: Quantifying Vocabulary Mismatch in E-Commerce Site Search
searchhub.io·6h·
Discuss: Hacker News
📈Search Quality
Context Length Alone Hurts LLM Performance Despite Perfect Retrieval
arxiv.org·1d
🧠Intelligence Compression
Evolutionary Profiles for Protein Fitness Prediction
arxiv.org·16h
🔗Functional Compilers
H1B-KV: Hybrid One-Bit Caches for Memory-Efficient Large Language Model Inference
arxiv.org·1d
💨Cache Optimization
Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management
arxiv.org·16h
💻Local LLMs