"Not sure about new training methods" -> they said the following in their system card (
https://
cdn.openai.com/gpt-4-5-system
-card.pdf
…): "We trained it using new
supervision techniques combined with traditional methods"
MACHINE LEARNING
-
GPT-4.5 trained with new supervision techniques, OpenAI confirms
By
–
-
Google Paper on Machine Learning from Summer 2024
By
–
Also reminds me of this nice paper by Google last summer: https://
arxiv.org/abs/2408.03314 -
Train vs Inference Compute Scaling in LLMs: An Orthogonal Comparison
By
–
I think that's an apple-to-oranges comparison. We don't now how GPT4.5 looks like with inference-compute scaling. Train- and inference-compute are two orthogonal ways to improve LLMs.
-
The Growing Role of Artificial Intelligence in Education
By
–
The Growing Role Of Artificial Intelligence In The Educational Sector
#AI #AIio #BigData #ML #NLU #Futureofwork @lexfridman @sama @kaifulee @ID_AA_Carmack @karpathy @2morrowknight @ylecun -
Synthetic Data Transforms AI/ML: Access, Quality, Privacy
By
–
"There's a real-world demonstration of how invaluable synthetic data is." @yevmeyer
, Chief Scientist at @gretel_ai
, shares how synthetic data transforms AI/ML by improving data access, quality, privacy and model training. Watch the full episode: https://
dbricks.co/4hVAzeE -
Deploy Your Own AI Model on Cloud with Open-Source Tools
By
–
What a week! Tired of reading about other companies' AI deployments? Here's a nice hands-on tutorial to help you deploy your own AI model on a public or private cloud built on open-source tools for a change. https://t.co/8tjokHjnk9
— Sebastian Raschka (@rasbt) 28 février 2025What a week! Tired of reading about other companies' AI deployments? Here's a nice hands-on tutorial to help you deploy your own AI model on a public or private cloud built on open-source tools for a change.
-
GPT-4.5 Development: ML and Systems Engineering Challenges
By
–
it's very hard to get the math and ML right on a run as big as GPT-4.5, and requires difficult work at the intersection of ML and systems. @ColinWei11 , Yujia Jin, and @MikhailPavlov5 did excellent work to make this happen!
-
Latest AI Papers: SWE-RL and LoRA Improvements
By
–
Since you all asked for this list: – 25 Feb, SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution, https://
arxiv.org/abs/2502.18449
– 24 Feb, Make LoRA Great Again: Boosting LoRA with Adaptive Singular Values and Mixture-of-Experts Optimization Alignment, -
Token increase, parameters, and training methods improvements
By
–
sure, but due to more tokens, more parameters, new training methods, or all of the above 😛
-
New Supervision Techniques Combined with SFT and RLHF Training
By
–
So, it's some new sort "supervision techniques" > "We trained it using new supervision techniques combined with traditional methods like supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), similar to those used for GPT-4o."