When Reinforcement is chosen, users will be able to specify a Grarer schema. There is also an option to tell OpenAI how you want it to be graded, and it will generate the schema for you.
@testingcatalog
-
OpenAI offers 3 fine-tuning methods
By
–
OpenAI Platform Alpha users will be able to select from 3 different fine-tuning methods: – Supervised
– Direct Reference Optimisation
– Reinforcement Reinforcement fine-tuning with o1-mini has been presented today during the OpenAI Day 2 -
Gemini vs Claude creativity comparison
By
–
Interesting – it is better than before, Gemini looks quite creative. But Claude is still the best at this
-
Meta launches Llama 3.3 70B, more efficient and cheaper
By
–
Meta released Llama 3.3 70B, which matches the level of the latest 405B model at a lower cost. Already available on @GroqInc
-
New Google Gemini-exp-1206 2M context model
By
–
BREAKING : Google released a new Gemini-exp-1206 2M context model on AI Studio. This model has been in testing on lmarena for some time and now it is capturing its top charts. Looks really promising
-
OpenAI alpha application form for research program
By
–
Here is an alpha application form https://
openai.com/form/rft-resea
rch-program/
… -
OpenAI adds o1-mini to fine-tuning
By
–
Teasing o1-mini in the fine-tuning dropdown on OpenAI Platform
-
A legal agent fine-tuned with o1-mini?
By
–
Will we see an o1-mini fine-tuned legal agent? The model will learn how to reason in a custom domain
-
Developers can fine-tune o1 for custom LLMs
By
–
Fine-tuning will be possible for o1. Devs will be able to build o1 level custom use LLMs
-
ChatGPT ‘All Tools’ deployed for some users
By
–
ChatGPT 'All Tools' started rolling out to some users. This feature will allow you to use Canvas, Image gen and other 4o capabilities with the same model. h/t @curiousgangsta