Some notes on swapping out the "important" parts of neural nets
Quazim0t0
• • 2NEW Articles from Team or Enterprise organizations will get promoted to the main section.
CeleBERTy Small: Domain model for Pop Culture, Art, Music and Entertainment
NeuML
• • 3tuetoken: a faster drop-in tokenizer for LLMs
vdmbrsv
• • 1The Hidden Bottleneck in Production AI Isn't the Model — It's Context Management
arybach
• • 1"Really Great Transformer Article"
nroggendorff
• • 1AlphaFold2 vs ESMFold on the Mosaic Q law (Chapter I: human thioredoxin)
ProteinsMosaicQProject
• A brief history of distillation in AI
sergiopaniego
• • 4Two Silent Traps in Agentic LLM Evaluation: Vanishing Tool Calls and Disagreeing Judges (Lessons from a DGX Spark Quantization Study)
GurkanOz
• Your Fine-Tune Survives 4-Bit: Trained Agentic Skills Are More Quantization-Robust Than Knowledge
GurkanOz
• AI Companies in India: Top AI, GenAI, Agentic AI & Listed AI Companies in 2026
daya-shankar
• 🔁 Apprendre à un LLM français de 15M à penser plus profond — et à savoir quand s'arrêter 🇫🇷
RDTvlokip
• • 3🔁 Teaching a 15M French LLM to think deeper — and to know when to stop 🇫🇷
RDTvlokip
• • 3Adding a GPU Without Building One
FINAL-Bench
• • 17CiberIA Expected Cognitive Profile Fast: AMALIA-9B-0625-SFT
gcjordi
• How I Built an AI Scholarship Finder with Olostep
kingabzpro
• • 2Catching Confident Hallucinations from the Inside
assafpet
• • 2Hidden-State Hallucination Probes Beat Confidence Baselines — And the Gap Grows with Scale
sarimahsan101
• CiberIA Expected Cognitive Profile Fast: Mythos - Fable
gcjordi
• Claude Fable 5 — Technical Harness Report
Svngoku
• • 5Pulpie: Pareto-Optimal Models for Cleaning the Web
feyninc
• • 6