Han
@hanchunglee.bsky.social
Pinned
Han
@hanchunglee.bsky.social
· Nov 25
Reasoning Series, Part 4: Reasoning with Compound AI Systems and Post-Training
Explore how compound AI systems and post-training approaches can make large language models (LLMs) more reliable and scalable by improving their reasoning capabilities. Learn about validation, verific...
leehanchung.github.io
🧠 New blog post: Reasoning with Compound AI Systems and Post-Training
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
📊 Want to level up your AI/ML evaluations?
Learn how Bootstrap Resampling can add confidence intervals to your metrics. Perfect for tasks like classification & LLM-based scoring!
📖 Read more:
leehanchung.github.io/blogs/2024/1...
Learn how Bootstrap Resampling can add confidence intervals to your metrics. Perfect for tasks like classification & LLM-based scoring!
📖 Read more:
leehanchung.github.io/blogs/2024/1...
Statistics for AI/ML, Part 1: Adding Confidence Interval to Your Aggregation Statistics
This blog post explores a key intersection of statistics and AI/ML evaluation metrics, focusing on how to add confidence intervals to aggregation statistics using bootstrap resampling. Whether you're ...
leehanchung.github.io
December 25, 2024 at 5:49 AM
📊 Want to level up your AI/ML evaluations?
Learn how Bootstrap Resampling can add confidence intervals to your metrics. Perfect for tasks like classification & LLM-based scoring!
📖 Read more:
leehanchung.github.io/blogs/2024/1...
Learn how Bootstrap Resampling can add confidence intervals to your metrics. Perfect for tasks like classification & LLM-based scoring!
📖 Read more:
leehanchung.github.io/blogs/2024/1...
open source open source.
Releasing: a dataset of two million Bluesky posts.
This dataset has been collected using Bluesky's API, and I hope it will be useful for all the researchers out there!
This dataset has been collected using Bluesky's API, and I hope it will be useful for all the researchers out there!
November 27, 2024 at 10:06 PM
open source open source.
🧠 New blog post: Reasoning with Compound AI Systems and Post-Training
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
Reasoning Series, Part 4: Reasoning with Compound AI Systems and Post-Training
Explore how compound AI systems and post-training approaches can make large language models (LLMs) more reliable and scalable by improving their reasoning capabilities. Learn about validation, verific...
leehanchung.github.io
November 25, 2024 at 3:48 PM
🧠 New blog post: Reasoning with Compound AI Systems and Post-Training
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
Dive into making LLMs reason better through:
- verification
- majority voting
- reasoning tools
- search
- post-training for reasoning🤔
leehanchung.github.io/blogs/2024/1...
ml anon!!
So first version of an ml anon starter pack. go.bsky.app/VgWL5L Kept half-anons (like me and Vic). Not all anime pfp, but generally drawn.
November 24, 2024 at 7:14 PM
ml anon!!
hmm what’s the best way to bookmark on this site, or to have different groups to follow, e.g., tweetdeck
November 21, 2024 at 11:12 PM
hmm what’s the best way to bookmark on this site, or to have different groups to follow, e.g., tweetdeck
ACL is not an RL conference.
November 20, 2024 at 10:44 PM
ACL is not an RL conference.
you don’t have a service mesh. you have a clock mesh.
Company: We have a monolith!
Me: ...
Company: *holds up diagram of 8 services, 15 databases, and a home grown queue implementation*
Me: You fucked up a perfectly good distributed system is what you did. Look at that thing, it's got clock skew.
Me: ...
Company: *holds up diagram of 8 services, 15 databases, and a home grown queue implementation*
Me: You fucked up a perfectly good distributed system is what you did. Look at that thing, it's got clock skew.
November 20, 2024 at 2:00 PM
you don’t have a service mesh. you have a clock mesh.
Reposted by Han
Even as an interpretable ML researcher, I wasn't sure what to make of Mechanistic Interpretability, which seemed to come out of nowhere not too long ago.
But then I found the paper "Mechanistic?" by
@nsaphra.bsky.social and @sarah-nlp.bsky.social, which clarified things.
But then I found the paper "Mechanistic?" by
@nsaphra.bsky.social and @sarah-nlp.bsky.social, which clarified things.
November 20, 2024 at 8:00 AM
Even as an interpretable ML researcher, I wasn't sure what to make of Mechanistic Interpretability, which seemed to come out of nowhere not too long ago.
But then I found the paper "Mechanistic?" by
@nsaphra.bsky.social and @sarah-nlp.bsky.social, which clarified things.
But then I found the paper "Mechanistic?" by
@nsaphra.bsky.social and @sarah-nlp.bsky.social, which clarified things.
Reposted by Han
From double descent to grokking, deep learning sometimes works in unpredictable ways.. or does it?
For NeurIPS(my final PhD paper!), @alanjeffares.bsky.social & I explored if&how smart linearisation can help us better understand&predict numerous odd deep learning phenomena — and learned a lot..🧵1/n
For NeurIPS(my final PhD paper!), @alanjeffares.bsky.social & I explored if&how smart linearisation can help us better understand&predict numerous odd deep learning phenomena — and learned a lot..🧵1/n
November 18, 2024 at 7:25 PM
From double descent to grokking, deep learning sometimes works in unpredictable ways.. or does it?
For NeurIPS(my final PhD paper!), @alanjeffares.bsky.social & I explored if&how smart linearisation can help us better understand&predict numerous odd deep learning phenomena — and learned a lot..🧵1/n
For NeurIPS(my final PhD paper!), @alanjeffares.bsky.social & I explored if&how smart linearisation can help us better understand&predict numerous odd deep learning phenomena — and learned a lot..🧵1/n
is this the second coming of butterfly?
November 1, 2024 at 12:06 AM
is this the second coming of butterfly?
Is it really dying
Bird app dying
October 26, 2024 at 1:16 PM
Is it really dying
Network discovery here is kinda slow. :(
May 8, 2023 at 7:16 AM
Network discovery here is kinda slow. :(