For example, you can give TestingCatalog URL and explore the content via NotebookLM. It feels like it is trying to scrape the date up until it reaches a certain context window limit. So you never know how complete and reliable it would be
LLMS
-
Google NotebookLM adds URL support for source material
By
–
Google's NotebookLM now can take any URL as a source. It would let you chat and take notes based on the conversation with the date from this source.
-
Meta AI introduces vision-language models educational paper
By
–
New from Meta AI Want to learn more about VLMs? Ask Florian Bordes from @AIatMeta about their latest paper “An Introduction to Vision-Language Models”. This week’s paper of the week https://
alphaxiv.org/abs/2405.17247 -
Etymology and Tokenization in Language Models Explained
By
–
Ah this makes sense. So etymology is baked into the tokenization in this way. I guess there's similarity to phoenetic languages in that sense.
-
72B Model Benchmarking: MixEval and MMLU-Pro Performance Analysis
By
–
The 72B model looks amazing but wow what an excellent benchmarking! Super happy to see MixEval and MMLU-Pro there. I have high hopes for MixEval as a proxy for Arena Elo. Also curious to see if these models behave better than Llama 3 for SFT.
-
User habits in managing AI chat history
By
–
This would include a sorting option for used or created chat sorting. Do you revisit your past conversations often or just create a new chat like I do?
-
OpenAI publishes new research to understand GPT-4 internals
By
–
OpenAI has come under fire for letting its superalignment team disappear but the company just published work from @ilyasut
, @janleike
, and others on a new way to peer inside GPT-4 to better understand how it works. -
GPT-5 Release Announcement Surprises AI Community
By
–
Wait, WHAT DE FUQ? https://
r.mtdv.me/blog/posts/gpt
-5
… -
SAE for LLM Feature Extraction and Behavior Conditioning
By
–
Y lo que ha sacado OpenAI, pues bueno, se le parece demasiado… → Explorar el uso de SAE.
→ Para extraer features de un LLM.
→ Que podrían servir para condicionar el comportamiento del modelo. Ajam -
Anthropic’s Scaling Monosemanticity: Breakthrough in Transformer Interpretability
By
–
Es el tema del próximo vídeo y es FAS-CI-NAN-TE, pero el crédito va a Anthropic que no sólo ha hecho un increíble trabajo de interpretabilidad, sino también de documentación que os invito a leer si os interesa. https://
transformer-circuits.pub/2024/scaling-m
onosemanticity/index.html
…