Here you go: https://
arxiv.org/abs/2401.02994
MACHINE LEARNING
-
Chai Research Hiring AI/ML Roles Up to $500k
By
–
Chai Research is a Generative AI Company that serves 15 trillion+ tokens every month close to the scale of Anthropic AI with just a team of 12. Best part? They are hiring for AI/ML roles ranging from $100k to $500k.
-
Blending Smaller AI Models for Cost-Effective Performance
By
–
Blending smaller AI model enables: 1. Cost-Effective Performance: Combining three mid-sized models (6B/13B parameters) can rival or surpass the capabilities of a single large model without the hefty computational demands.
-
Smaller AI Models Collaborate to Match ChatGPT Performance
By
–
In the race for better conversational AI, bigger isn’t always better. Recent AI research by @chai_research introduces “Blending,” a method where smaller AI models collaborate to match or even exceed the performance of massive models like ChatGPT (175B+ parameters).
-
Hermes 2 Θ: Open Source Language Model Innovation
By
–
This is a really cool new model, and a hint at the future of open source language models. Synthetic data, model merges and RLHF are becoming crucial tools in the AI hacker's toolkit. Try Hermes 2 Θ on Replicate right now at https://
replicate.com/nousresearch/h
ermes-2-theta-llama-8b
… and let us know what you think! -
IMO-Bench: Gemini Math Achieves 25% on Mathematical Olympiad Problems
By
–
Glad to see us keep pushing the frontier on maths! Also a bit of a teaser to IMO-Bench (after #alphageometry), which my team (with @quocleix
's support) built & the best model only got 25% 🙂 Check out the cool answer of the Gemini Math on an APMO problem in @OriolVinyalsML
's -
Model Evaluation and Reasoning-Focused Haystack Testing Improvements
By
–
I don’t always find the middle to be lost — really depends on the model/use-case. We definitely need better (more reasoning-focused) haystack tests!
-
Machine Learning: Correlation Over Causation in Practice
By
–
Machine learning is the art of doing the wrong things that correlate with the right ones.
-
Gemini 1.5 Pro Achieves Record 91.1% on MATH Benchmark
By
–
One other thing in the updated Gemini 1.5 Pro report: we show how a research model that is a mathematics-specialized version of 1.5 Pro achieves a record score of 91.1% on the MATH benchmark (the SOTA just 3 years ago, in May, 2021 was 6.9%!).
-
Extended Machine Translation Evaluation Results in Updated Report
By
–
The updated report is now 153 pages, and has quite a few new results. In the February report, I found the results on Kalamang translation for the Machine Translation from One Book benchmark quite exciting. In this updated report, we’ve extended this line of evaluation to test