Currently at the AI Summit Seoul in BTS' home country. Just after I finished my talk, the next speaker used the graph™. I can't escape it.
@maximelabonne
-
Sora’s Training Data Skewed Toward Skyrim Content
By
–
Too much Skyrim in Sora's pre-training mix pic.twitter.com/atYrJnRrke
— Maxime Labonne @ ICLR (@maximelabonne) 11 décembre 2024Too much Skyrim in Sora's pre-training mix
-
Llama 3.1 Tool Use Benchmarks Released
By
–
I'm not sure. They even released tool use benchmarks for Llama 3.1
-
Llama 3.3 70B Released: New Features and Language Support
By
–
Llama 3.3 70B – The new license doesn't change anything compared to Llama 3.1 – Supports 8 languages: English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai (no Chinese, Japanese?) – Added "Tool-use" but it was already there…?
-
BigLlama-3.1-1T-Instruct Model Issues Revealed
By
–
Remember BigLlama-3.1-1T-Instruct (
https://
huggingface.co/mlabonne/BigLl
ama-3.1-1T-Instruct
…)? *cough* Haha, awkward. -
STAR Evolution Optimizes 18 Generations Quality Size
By
–
And here are 18 generations with populations that are iteratively assessed, paired, recombined, and mutated with STAR evolution to optimize both quality and size.
-
STAR Framework Evolves AI Model Architectures
By
–
STAR: Synthesis of Tailored Architectures Really cool work from my colleagues @liquidai on model architecture. This framework creates populations of architectures and uses evolutionary algorithms to iteratively create better solutions.
-
LLMs in Medicine: Transparency and EU AI Act Implications
By
–
Happy to announce that our paper has been published in Nature's npj Digital Medicine We argue that LLMs in medicine must be transparent and controllable. We discuss the EU AI Act, which could potentially restrict responsible usage of LLMs. Paper: https://
nature.com/articles/s4174
6-024-01344-w
… -
Teaching Math to LLMs: New Career Opportunity Emerges
By
–
I just got offered a role teaching math to LLMs.
-
Comprehensive llama.cpp Guide: Building and Performance Benchmarking
By
–
Cool guide to llama.cpp from building to perf benchmarking llama.cpp's documentation is notoriously difficult to work with. There's some more advanced stuff I didn't know about, it's great stuff from @steel_ph0enix https://
steelph0enix.github.io/posts/llama-cp
p-guide/
…