SPIRALbase, Part 3: SPIRAL Cortex and Policy-First Memory
NoeticDiffusion
• NEW Articles from Team or Enterprise organizations will get promoted to the main section.
🧠 I trained my own French LLM from scratch — alone, with a 1080 Ti, and the power went out ⚡🇫🇷
RDTvlokip
• • 6🧠 J'ai entraîné mon propre LLM français from scratch — seul, avec une 1080 Ti, et le courant a coupé ⚡🇫🇷
RDTvlokip
• • 2AI Safety Without Panopticons: A Canadian Framework for Companion AI That Protects Users Without Turning Private Chats Into Police Files
Clock070303
• • 1Building MoulSot: How We Curated 1,500 Hours of Moroccan Darija Speech, Selected the Best 80 Hours for Transcription, and Fine-Tuned Qwen3-ASR on Top of It
abdeljalilELmajjodi
• • 2Organon: a skill-first agent architecture for end-to-end scientific workflows
kerdel
• • 1Pixel Art Bench: Evaluating Structured Generation in Language Models
AINovice2005
• • 1Build a legal RAG app that won't be held in contempt
isaacus
• • 4THE RANDOM CURSE <br> Train/Val split on CIFAR 10, PASCAL VOC and IMDB
Yann-CV
• • 1From Image to Music Language: Starry's Two-Stage Approach for Complex Polyphonic OMR
k-l-lambda
• Talking to a 4-Year-Old: A Multilingual Benchmark for Children's AI Companions
batuhanaktas
• • 4Feather DB on LongMemEval: embedded retrieval beats full-context GPT-4o for $2.40
Sri-Vigneshwar-DJ
• Building a Multi-Task English NLP Cascade in the Open
milliondreams
• Fused Batched Thin SVD, Part II: Extending the Jacobi Pipeline to N=6 with Configurable Convergence
AbstractPhil
• Under The Hood : Trinity-Large-Thinking Disected
balaatdell
• • 1Under The Hood : Cohere Transcribe Deep Dive
balaatdell
• Under The Hood : Kimi K2.5 Disected
balaatdell
• Agentic Scientific Machine Learning for Neural Operators
hugging-science
• • 2Training low-bit ternary models with Axolotl
axolotl-ai-co
• • 5ModernBERTic: The First Modern Encoder for South Slavic Languages
permitt
•