Lightnews — Scholar-powered news

Reposted by Martin Wattenberg

vrli.bsky.social @vrli.bsky.social · Aug 4

Charts and graphs help people analyze data, but can they also help AI?

In a new paper, we provide initial evidence that it does! GPT 4.1 and Claude 3.5 describe three synthetic datasets more precisely and accurately when raw data is accompanied by a scatter plot. Read more in🧵!

1 2 8

Reposted by Martin Wattenberg

Jonathan Zittrain @zittrain.bsky.social · May 21

AI is often thought of as a black box -- no way to know what's going on inside. That's changing in eye-opening ways. Researchers are finding "beliefs" models are forming as they converse, and how those beliefs correlate to what the models say and how they say it.

www.theatlantic.com/technology/a...

What AI Thinks It Knows About You

What happens when people can see what assumptions a large language model is making about them?

www.theatlantic.com

4 13 32

Reposted by Martin Wattenberg

Baby Name Wizard author Laura Wattenberg @babynames.bsky.social · May 12

The interactive NameGrapher is updated with 2024 baby name popularity stats! Come explore--and marvel that Oliver and Olivia have converged namerology.com/baby-name-gr...

Historical popularity chart showing the popularity of Oliver rising to meet the previously much greater popularity of Olivia

1 2 8

Martin Wattenberg @wattenberg.bsky.social · May 12

A wonderful visualization for those of us obsessed by sunlight and geography!

Olivia Vane @oliviafvane.bsky.social · May 12

This map shows the hour of sunrise globally through the year. It reveals time zones following national and, sometimes, regional boundaries, and slicing through the oceans.

1 1 27

Martin Wattenberg @wattenberg.bsky.social · Mar 27

An incredibly rich, detailed view of neural net internals! There are so many insights in these papers. And the visualizations of "addition circuit" features are just plain cool!

Chris Olah @colah.bsky.social · Mar 27

Can we understand the mechanisms of a frontier AI model?

📝 Blog post: www.anthropic.com/research/tra...
🧪 "Biology" paper: transformer-circuits.pub/2025/attribu...
⚙️ Methods paper: transformer-circuits.pub/2025/attribu...

Featuring basic multi-step reasoning, planning, introspection and more!

On the Biology of a Large Language Model

transformer-circuits.pub

2 16

Martin Wattenberg @wattenberg.bsky.social · Mar 27

Great news, congrats! And glad you’ll still be in the neighborhood!

1

Martin Wattenberg @wattenberg.bsky.social · Mar 24

I'd be curious about advice on teaching non-coders how to test programs they've written with AI. I'm not thinking unit tests so much as things like making sure you can drill down for verifiable details in a visualization—basic practices that are good on their own, but also help catch errors.

2 10

Martin Wattenberg @wattenberg.bsky.social · Mar 24

Now that we have vibe coding, we need vibe testing!

6 4 23

Martin Wattenberg @wattenberg.bsky.social · Mar 21

Oh, that looks super relevant and fascinating, reading through it now...

1

Martin Wattenberg @wattenberg.bsky.social · Mar 21

Ha! I think (!) that for me, the word "calculate" connotes narrow precision and correctness, whereas "think" is more expansive but also implies more fuzziness and the possibility of being wrong. That said, your observation does give me pause!

Martin Wattenberg @wattenberg.bsky.social · Mar 21

We're following the terminology of the DeepSeek-R1 paper that introduced this model: arxiv.org/abs/2501.12948 Whether it's really the best metaphor is certainly worth asking! I can see pros and cons for both "thinking" and "calculating"

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT)...

arxiv.org

1 1

Martin Wattenberg @wattenberg.bsky.social · Mar 21

These are great questions! I believe there's at least one graph of p(correct answer) on the main Arbor discussion page, and generally there are a lot more details: github.com/ARBORproject...

Reasoning or Performing: locating "breakthrough" in the model's reasoning · ARBORproject arborproject.github.io · Discussion #11

Research Question When asked the DeepSeek models a challenging abstract algebra question, they often generated hundreds of tokens of reasoning before providing the final answer. Yet, on some questi...

github.com

1 1

Martin Wattenberg @wattenberg.bsky.social · Mar 21

Interesting question! I haven't calculated this, but @yidachen.bsky.social might know

1

Martin Wattenberg @wattenberg.bsky.social · Mar 21

This is a common pattern, but we're also seeing some others! Here are similar views for multiple-choice abstract algebra questions (green is the correct answer; other colors are incorrect answers) You can see many more at yc015.github.io/reasoning-pr... cc @yidachen.bsky.social

Colorful depictions of reasoning progress: most of the time the system settles on the correct answer but sometimes it vacillates in interesting ways.

3 5

Martin Wattenberg @wattenberg.bsky.social · Mar 13

Very cool! You're definitely not alone in finding this fascinating. If you're looking for other people interested in this kind of thing, drop by the Arbor Project page, if you haven't already. github.com/ArborProject...

GitHub - ARBORproject/arborproject.github.io

Contribute to ARBORproject/arborproject.github.io development by creating an account on GitHub.

github.com

1 3

Martin Wattenberg @wattenberg.bsky.social · Mar 3

The wind map at hint.fm/wind/ has been running since 2012, relying on weather data from NOAA. We added a notice like this today. Thanks to @cambecc.bsky.social for the inspiration.

1 21 78

Martin Wattenberg @wattenberg.bsky.social · Feb 26

It's based on a data set of multiple-choice questions that have a known right answer, so this visualization only works when you have labeled ground truth. Definitely wouldn't shock me if those answers were labeled by grad students, though!

3

Martin Wattenberg @wattenberg.bsky.social · Feb 25

Great questions! Maybe it would be faster... or maybe it's doing something important under the hood that we can't see? I genuinely have no idea.

1 1

Martin Wattenberg @wattenberg.bsky.social · Feb 25

We also see cases where it starts out with the right answer, but eventually "convinces itself" of the wrong answer! I would love to understand the dynamics better.

1

Martin Wattenberg @wattenberg.bsky.social · Feb 25

You can see the model go down the wrong path, "realize" it's not right, then find the correct answer! To see more visualizations, or if you have related ideas, join the discussion here!
github.com/ARBORproject... (vis by @yidachen.bsky.social in conversation with @diatkinson.bsky.social )

Reasoning or Performing · ARBORproject arborproject.github.io · Discussion #11

Research Question When asked the DeepSeek Distilled R1 models a challenging abstract algebra question, they often generated hundreds of tokens of CoT before providing the final answer. Yet, on some...

github.com

1 11

Martin Wattenberg @wattenberg.bsky.social · Feb 25

Neat visualization that came up in the ARBOR project: this shows DeepSeek "thinking" about a question, and color is the probability that, if it exited thinking, it would give the right answer. (Here yellow means correct.)

6 15 81

Martin Wattenberg @wattenberg.bsky.social · Feb 22

Thank you! That's a great write-up, and this is definitely an interesting experiment. The distinction between how the model might do parsing vs. solving is very much worth thinking about. I added a few thoughts on the wiki page. github.com/ARBORproject...

Chain of Thought for Tsumego (Go Life or Death) Problems

Contribute to ARBORproject/arborproject.github.io development by creating an account on GitHub.

github.com

1 2

Martin Wattenberg @wattenberg.bsky.social · Feb 20

Great thread describing the new ARBOR open interpretability project, which has some fascinating projects already. Take a look!

David Bau @davidbau.bsky.social · Feb 20

ARBOR aims to accelerate the internal investigation of the new class of AI "reasoning" models.

See the ARBOR discussion board for a thread for each project underway.

github.com/ArborProjec...

2 8

Martin Wattenberg @wattenberg.bsky.social · Feb 20

Excellent idea! I just added an "observations" index page on the wiki for things like that.
github.com/ARBORproject...

Observations

Contribute to ARBORproject/arborproject.github.io development by creating an account on GitHub.

github.com

1 1

Martin Wattenberg @wattenberg.bsky.social · Feb 20

Take a look at some initial research projects, and see if there's one you'd like to work on:
github.com/ARBORproject...
Or propose your own idea! There are many ways to contribute, and we welcome all of them.

ARBORproject arborproject.github.io · Discussions

Explore the GitHub Discussions forum for ARBORproject arborproject.github.io. Discuss code, ask questions & collaborate with the developer community.

github.com

1 2 8