Just heard that Gemini Code Assist, Google's alternative to GitHub Copilot, is now free, which is fantastic news. Additionally, Google released a free GitHub code review agent! This is great, because it lowers the barrier to entry to try and incorporate these tools into our
@rasbt
-
Hugging Face R1 reproduction pipeline adopts uv tool
By
–
Was just trying out the @huggingface R1 reproduction pipeline and saw they are recommending and using uv now too https://
github.com/huggingface/op
en-r1
… -
Claude Version Naming Improvements Discussion
By
–
Not sure if 'Claude 3.5 Sonnet & new Claude 3.5 Sonnet' was that much better than 'o1-preview and o1' in terms of version naming. But yeah, glad they fixed that with 3.7
-
Parallel Best-of-N Sampling and Majority Voting for LLM Improvement
By
–
Sure! The simplest example would be parallel Best-of-N sampling and using majority vote or a verifier. Google had a good paper on that last summer: https://
arxiv.org/abs/2408.03314 -
Multi-head Latent Attention Innovation and OpenAI’s Mixture-of-Experts
By
–
Good point. They brought us multi-head latent attention last summer.
(I'm not counting mixture-of-experts, because OpenAI uses it regardless of constraints :P) -
Apple Intelligence Implementation Based on Official Paper
By
–
Yes, Apple Intelligence uses it if we can believe Apple's paper (
https://
arxiv.org/abs/2407.21075): -
LLM thinking toggle routing and model selection strategies
By
–
Interesting. Maybe it's one with thinking always-on and one always-off. Similar to the IBM Granite model with the toggle. And on the Claude website, maybe they have another "small" LLM routing the user prompt / controlling that toggle.
-
Inference-Time Scaling: Prompt Modification vs Sampling and Verification
By
–
*One might argue that modifying prompts (eg classic "think step by step") counts as inference-time scaling due to the extra output tokens. So in that sense they use inference-time scaling.
I should have been more clear: I meant specifically sampling techniques + using a verifier. -
Compute Constraints Led to LoRA and DPO Development
By
–
Compute constraints brought us LoRA and DPO
-
System Card Analysis: Reasoning Implementation via RL Toggle
By
–
What a week! Just read the system card, and it looks like they implemented reasoning via RL. My guess is the thinking on/off toggle is likely a system prompt. I wonder if they added inference-time scaling like o1 or if it’s just RL like R1. Anyone found details on that?