thanks for sharing — seems like this is some odd student-teacher setting though; i'm just training a single model with cross-entropy
AI
-
Disagreement on Prompt Writing Style and Natural Language Coding
By
–
Many of my prompts are a bit more like writing code in natural langauge (english) — so I've got to disagree. Still, I appreciate you sharing a dissenting view.
-
Belief that consistent prompting improves overall clarity of thought
By
–
Absolutely. I believe that after a couple years of prompting, I actually think more clearly overall.
-
Prompt-Writing Apps Raise the Floor But Understanding Matters Most
By
–
I've tested some of these prompt-writing apps. They're starting to do a pretty good job. I've even used them as the basis for my own prompts from time-to-time. But if you don't understand *why* they are better, you'll only ever get so far with them. I think they raise the floor
-
Contrastive Learning with Large Batches for Embedding Models
By
–
sort of; i'm doing contrastive learning with large batches; at the end it's computed as cross-entropy, but this isn't a typical language modeling setup, i'm training an embedding model
-
Prompt Engineering: Waste of Time or Secret Weapon?
By
–
Prompt engineering: Waste of time or secret weapon? In Co-Intelligence, Ethan Mollick effectively argues against "complex" prompt engineering. He rarely spends much time coaxing better output from resistant AI. But here's the thing… I've found that using the same
-
Stanford AI Index Report 2026 Launches Next Week
By
–
Waiting for the next issue of the AI Index? Be sure to subscribe to @StanfordHAI
’s newsletter to get this year’s report when it goes live next week on Monday, April 15: -
Cosine Decay Implementation Bug in Training Loop
By
–
it's supposed to be cosine decay but this is just linear warmup and my thing is already broken
-

Snorkel Custom: Enterprise LLM Application Accelerator Announced
By
–
Today we're excited to announce Snorkel Custom, a hands-on accelerator for enterprises using LLMs to build AI applications. It pairs Snorkel Flow, an AI data development platform, with Snorkel AI researchers and ML practitioners.
Learn more: https://
snkl.ai/custom -

Model Degradation Without Gradient Norm Increase Troubleshooting
By
–
model-trainers of Twitter: why does my model start getting worse over time without the grad norm noticeably increasing — how can this be possible? and how do i fix it? (again, please help 🙂 )