💻 Local LLMs

Model Quantization, Inference Optimization, GGUF Format, Privacy-preserving AI

ChatGPT and other AI models can be ‘poisoned’ to spew gibberish, researchers warn
the-independent.com·10h
🔓Hacking
LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
arxiv.org·22h
🔲Cellular Automata
Integral Signatures of Activation Functions: A 9-Dimensional Taxonomy and Stability Theory for Deep Learning
arxiv.org·22h
🧠Machine Learning
In-Depth Analysis: "Attention Is All You Need"
dev.to·11h·
Discuss: DEV
🧠Intelligence Compression
Mitigating Judgment Preference Bias in Large Language Models through Group-Based Polling
arxiv.org·22h
🧠Intelligence Compression
How to Teach Large Multimodal Models New Skills
arxiv.org·22h
📊Learned Metrics
Own your AI: Learn how to fine-tune Gemma 3 270M and run it on-device
developers.googleblog.com·2d·
Discuss: Hacker News
🌀Brotli Internals
Memory Retrieval and Consolidation in Large Language Models through Function Tokens
arxiv.org·22h
💻Programming languages
Self-Improving LLM Agents at Test-Time
arxiv.org·22h
🧠Intelligence Compression
Distilling Reasoning into Student LLMs: Local Naturalness for Selecting Teacher Data
arxiv.org·3d
🧮Kolmogorov Complexity
Exponential Error Bounds for Information Bottleneck Source Coding Problems
arxiv.org·22h
📐Compression Bounds
MARC: Memory-Augmented RL Token Compression for Efficient Video Understanding
arxiv.org·22h
🧠Learned Codecs
Why LLMs cannot reach GenAI, but why it looked like they could
haversine.substack.com·4h·
Discuss: Substack
🧠Intelligence Compression
Activation Alchemist: Sculpting Stability with Functional Signatures
dev.to·6h·
Discuss: DEV
🔍Concolic Testing
Learning from Failures: Understanding LLM Alignment through Failure-Aware Inverse RL
arxiv.org·2d
🧠Intelligence Compression
94% of Developers Waste Tokens on Reasoning LLMs. Here's Why.
dev.to·20h·
Discuss: DEV
🧮Z3 Solver
Optimal Stopping in Latent Diffusion Models
arxiv.org·22h
🧠Machine Learning
Aligning Large Language Models via Fully Self-Synthetic Data
arxiv.org·1d
🔗Monadic Parsing
Distribution Preference Optimization: A Fine-grained Perspective for LLM Unlearning
arxiv.org·3d
🧠Machine Learning
Revisiting Mixout: An Overlooked Path to Robust Finetuning
arxiv.org·1d
🧠Learned Codecs