Tweet

@HuaxiuYaoML

https://twitter.com/HuaxiuYaoML/status/1478243699771412481

This Thread may be Removed Anytime!

Twitter may remove this content at anytime! Save it as PDF for later use!

More from @chelseabfinn

Chelsea Finn

@chelseabfinn

Feb 9

What should ML models do when there's a *perfect* correlation between spurious features and labels?

This is hard b/c the problem is fundamentally _underdefined_

DivDis can solve this problem by learning multiple diverse solutions & then disambiguating
arxiv.org/abs/2202.03418
🧵

https://twitter.com/kchonyc/status/1455589673229770757

Prior works have made progress on robustness to spurious features but also have important weaknesses:
- They can't handle perfect/complete correlations
- They often need labeled data from the target distr. for hparam tuning

https://twitter.com/kchonyc/status/1455589673229770757

DivDis can address both challenges, using 2 stages:
1. The Diversify stage learns multiple functions that minimize training error but have differing predictions on unlabeled target data
2. The Disambiguate stage uses a few active queries to identify the correct function

Read 7 tweets

Chelsea Finn

@chelseabfinn

Oct 22, 2021

@_eric_mitchell_

Large language models (LLMs) often make mistakes that are difficult to correct.

We study the problem of quickly editing these models:
Paper: arxiv.org/abs/2110.11309
Code: github.com/eric-mitchell/…

w/ @_eric_mitchell_, C. Lin, @ABosselut, @chrmanning

thread 🧵👇

We assume a pre-trained model & a dataset that covers many possible model edits

Then, we meta-train a model editor that predicts a model update that:
- edits the model
- otherwise keeps the model behavior the same

(2/4)

You can train model editors for massive models (e.g. GPT-J, T5-11B) in <1 day on a single GPU.

Edits with the resulting model editor are extremely fast, with edit success rate of 80-90%.
(3/4)

Read 4 tweets

Chelsea Finn

@chelseabfinn

Sep 22, 2021

RL methods so often learn from _scratch_. Can they leverage offline experience from previous tasks?

They can. And if they do, they will learn new tasks ~2x faster.

Paper: arxiv.org/abs/2109.09180
Website: sites.google.com/view/retain-ex…

Led by Annie Xie. 🧵👇(1/4)

Many prior transfer learning methods try to transfer weights, e.g through fine-tuning.

We consider whether we can also transfer past *experiences*, rather than throwing away the prior data.
(2/4)

Retaining & filtering experiences performs *substantially* better than fine-tuning & other prior methods

It also outperforms learning from scratch, even when learning from scratch with 2x the data.
(3/4)

Read 4 tweets

Chelsea Finn

@chelseabfinn

Jul 20, 2021

@mike_h_wu

Thrilled to share new work on AI for education: can we give detailed, high-quality feedback to students?

Post: ai.stanford.edu/blog/prototran…
NYT Coverage: nytimes.com/2021/07/20/tec…

A collab w. the amazing @mike_h_wu @chrispiech & co 🧵

2/ Student feedback is a fundamental problem in scaling education.

Providing good feedback is hard: existing approaches provide canned responses, cryptic error messages, or simply provide the answer.

3/ Providing feedback is also hard for ML: not a ton of data, teachers frequently change their assignments, and student solutions are open-ended and long-tailed.

Supervised learning doesn’t work. We weren’t sure if this problem can even be solved using ML.

Read 12 tweets

Chelsea Finn

@chelseabfinn

Apr 1, 2021

@_anniechen_

How can robots generalize to new environments & tasks?

We find that using in-the-wild videos of people can allow learned reward functions to do so!
Paper: arxiv.org/abs/2103.16817

Led by @_anniechen_, @SurajNair_1
🧵(1/5)

To get reward functions that generalize, we train domain-agnostic video discriminators (DVD) with:
* a lot of diverse human data, and
* a narrow & small amount of robot demos

The idea is super simple: predict if two videos are performing the same task or not.
(2/5)

This discriminator can be used as a reward by feeding in a human video of the desired task and a video of the robot’s behavior.

We use it by planning with a learned visual dynamics model.
(3/5)

Read 5 tweets

Chelsea Finn

@chelseabfinn

Jul 8, 2020

@allan_zhou1

Convolution is an example of structure we build into neural nets. Can we _discover_ convolutions & other symmetries from data?

Excited to introduce:
Meta-Learning Symmetries by Reparameterization
arxiv.org/abs/2007.02933

w/ @allan_zhou1 @TensorProduct @StanfordAILab
Thread👇

To think about this question, we first look at how equivariances are represented in neural nets.

They can be seen as certain weight-sharing & weight-sparsity patterns. For example, consider convolutions.
(2/8)

We reparametrize a weight matrix into a sharing matrix & underlying filter parameters

It turns out this can provably represent any equivariant structure + filter parameters, for all group-equivariant convolutions with finite groups.
(3/8)

Read 8 tweets

Share this page!

Chelsea Finn

People who liked this thread also liked...

Try unrolling a thread yourself!

More from @chelseabfinn

Chelsea Finn

Chelsea Finn

Chelsea Finn

Chelsea Finn

Chelsea Finn

Chelsea Finn

Did Thread Reader help you today?

Don't want to be a Premium member but still want to support us?