Training a Helpful and Harmless Assistant with
Reinforcement Learning from Human Feedback slides: https://
docs.google.com/presentation/d
/1hYPWiLETSK5r_Y6sU0mWbOyAz-uevNJh3Xp6BmUnGHM/edit?usp=sharing
… paper: https://
arxiv.org/abs/2204.05862 instructgpt:
LLMS
-
Training Helpful Harmless Assistant with Reinforcement Learning from Human Feedback
By
–
-
Early RLHF Paper: From Supervised Fine-tuning to Personality in Language Models
By
–
[Slides] This is one of the earliest papers on RLHF (if not the first, alongside InstructGPT). Before RLHF, language models didn’t really have personalities—they mostly relied on supervised fine-tuning or clever prompting to understand humans. Think back to the InstructGPT days.
-
7 Alternative Attention Mechanism Implementations for Chapter 3
By
–
Welcome back, again :). I don't want to ruin your weekend, but based on that chapter, here are 7 more alternative ways to implement the attention mechanism https://
github.com/rasbt/LLMs-fro
m-scratch/blob/main/ch03/02_bonus_efficient-multihead-attention/mha-implementations.ipynb
…
(It's just for fun, pls don't let it hold you back from Ch 4!) -
Test-Time Compute Jargon Alert and Appreciation
By
–
Glad you found it helpful! Haha, and big-time jargon alert for "test-time compute"
-
Claude 3.7 Internal Testing: Team Feedback on AI Model Capabilities
By
–
We've been putting Claude 3.7 through its paces internally @every for the last few weeks Here's a vibe check from the whole team on how we use it and what it's good for:
-
Reasoning Emergence in LLMs Pre-training and Q&A Data
By
–
You mean that reasoning may emerge from pre-training itself? I think that current-gen LLMs are capable of some reasoning w/o explicit reasoning methodology applied to them. But then, the tricky part is most LLMs nowadays have Q&A data in their pre-training set. And if you think
-
MCP Protocol: Temporary Trend or Permanent AI Standard?
By
–
MCP – flash in the pan or future standard? Lots of buzz around MCP. @hwchase17 and @nfcampos debate whether it's here for the long run. Covers: – use cases for MCP
– Comparison to OpenAI Plugins
– Limitations of MCP Read the debate: https://
blog.langchain.dev/mcp-fad-or-fix
ture/
… Vote below: -
LLM Prompt Classification and Reasoning Toggle Strategies
By
–
Yeah, it's going to be interesting. You can do it either with 1) a 2nd cheap LLM that classifies the prompt as "reasoning" / "non-reasoning" and then modifies the system prompt and/or adds a special token as a toggle
2) do RLFH-style preference tuning to teach the LLM when to -
ChatGPT Writing Quality: AI Still Struggles With Voice
By
–
Been playing with ChatGPT to see how well it does authoring in my voice. tl;dr AI is still a poor writer that produces flat, anodyne writing. Still, I'm impressed by its ability to parse and synthesize large data sets and summarize it.
-
Choosing the Right AI Tool for Every Task
By
–
The most valuable lesson? Knowing which tool to use for which job. Sometimes you need Townie AI's app-building capabilities, sometimes Napkin AI's visualization strength, and other times Groq's raw speed. It's about the right tool for the right task