Author | Lightnews

Andrew Jesson @anndvision.bsky.social · Apr 23

if you’re at ICLR and also left your poster printing to the last minute

Colour Connect Pte Lte is goated

large format
1hr
$28
crispy

Andrew Jesson @anndvision.bsky.social · Apr 23

i'm at ICLR this week

presenting at the morning poster session on thursday

excited to catch up with friends and collaborators, old and new

let's chat

1 1

Andrew Jesson @anndvision.bsky.social · Dec 13

thanks to @yaringal.bsky.social , John P. Cunningham , and David Blei for their help !

1

Andrew Jesson @anndvision.bsky.social · Dec 13

fun @bleilab.bsky.social x oatml collab

come chat with Nicolas , @swetakar.bsky.social , Quentin , Jannik , and i today

Nicolas Beltran-Velez @velezbeltran.bsky.social · Dec 12

Hello!

We will be presenting Estimating the Hallucination Rate of Generative AI at NeurIPS. Come if you'd like to chat about epistemic uncertainty for In-Context Learning, or uncertainty more generally. :)

Location: East Exhibit Hall A-C #2703
Time: Friday @ 4:30
Paper: arxiv.org/abs/2406.07457

1 1 10

Andrew Jesson @anndvision.bsky.social · Dec 13

thank you to my co-authors @velezbeltran.bsky.social and @bleilab.bsky.social

Andrew Jesson @anndvision.bsky.social · Dec 13

we explore two different discrepancies: the negative log likelihood (NLL), and the negative log marginal likelihood (NLML)

the NLL gives p-values that are informative of whether there are enough in-context examples

this can reduce risk in safety critical settings

1

Andrew Jesson @anndvision.bsky.social · Dec 13

we show that the GPC is an effective OOD predictor on generative image completion tasks using a modified Llama-2 model trained from scratch

1

Andrew Jesson @anndvision.bsky.social · Dec 13

we show that the GPC is an effective predictor of out-of-capability natural language tasks using pre-trained LLMs

1

Andrew Jesson @anndvision.bsky.social · Dec 13

we show that the GPC is an effective OOD predictor for tabular data using synthetic data and a modified Llama-2 model trained from scratch

1

Andrew Jesson @anndvision.bsky.social · Dec 13

the result is the generative predictive p-value

pre-selecting a significance level α to threshold the p-value gives us a predictor of model capacity: the generative predictive check (GPC)

1

Andrew Jesson @anndvision.bsky.social · Dec 13

problem:

not all generative models (eg, LLMs) lend access to the likelihood and posterior

solution:

we can sample dataset completions from the predictive to simulate sampling from the posterior

and we can estimate the likelihood by conditioning on the completions

1

Andrew Jesson @anndvision.bsky.social · Dec 13

understanding these nuances is the domain of Bayesian model criticism

posterior predictive checks form a family of model criticism techniques

but for discrepancy functions like the negative log likelihood, PPCs require the likelihood and posterior

1

Andrew Jesson @anndvision.bsky.social · Dec 13

the posterior is informative about if there are enough in-context examples

but such inferences are made by any model, even misaligned ones

if a model is too flexible, more examples may be needed to specify the task

if it is too specialized, the inferences may be unreliable

1

Andrew Jesson @anndvision.bsky.social · Dec 13

a model θ defines a joint distribution over datasets x and explanations f

the joint comprises the likelihood over datasets and the prior over explanations

the posterior is a distribution over explanations given a dataset

the posterior predictive gives the model a voice

1

Andrew Jesson @anndvision.bsky.social · Dec 13

an in-context learning problem comprises a model, a dataset, and a task

knowing when an LLM provides reliable responses is challenging in this setting

there may not be enough in-context examples to specify the task

or the model may just not have the capability to it

1

Andrew Jesson @anndvision.bsky.social · Dec 13

can generative ai solve your in-context learning problem ?

we develop a predictor that only requires sampling and log probs

we show it works for tabular, natural language, and imaging problems

come chat at the safe generative ai workshop at NeurIPS

📄 arxiv.org/abs/2412.06033

1

Reposted by Andrew Jesson

Nicolas Beltran-Velez @velezbeltran.bsky.social · Dec 12

Hello!

We will be presenting Estimating the Hallucination Rate of Generative AI at NeurIPS. Come if you'd like to chat about epistemic uncertainty for In-Context Learning, or uncertainty more generally. :)

Location: East Exhibit Hall A-C #2703
Time: Friday @ 4:30
Paper: arxiv.org/abs/2406.07457

4 22

Andrew Jesson @anndvision.bsky.social · Dec 12

does your circuit preserve model behavior ?

does removing it disable the task ?

does it contain redundant parts ?

don ' t know ?

then come chat about hypothesis testing for mechanistic interpretability

poster 2803 East 4:30-7:30pm at NeurIPS

claudia shi @claudiashi.bsky.social · Dec 10

The circuit hypothesis proposes that LLM capabilities emerge from small subnetworks within the model. But how can we actually test this? 🤔

joint work with @velezbeltran.bsky.social @maggiemakar.bsky.social @anndvision.bsky.social @bleilab.bsky.social Adria @far.ai Achille and Caro

Reposted by Andrew Jesson

claudia shi @claudiashi.bsky.social · Dec 10

The circuit hypothesis proposes that LLM capabilities emerge from small subnetworks within the model. But how can we actually test this? 🤔

joint work with @velezbeltran.bsky.social @maggiemakar.bsky.social @anndvision.bsky.social @bleilab.bsky.social Adria @far.ai Achille and Caro

2 6 15

Andrew Jesson @anndvision.bsky.social · Nov 26

ofc
i love openreview

but damn
that ' s a lot of tabs

1

Reposted by Andrew Jesson

Sweta Karlekar @swetakar.bsky.social · Nov 20

(Shameless) plug for David Blei's lab at Columbia University! People in the lab currently work on a variety of topics, including probabilistic machine learning, Bayesian stats, mechanistic interpretability, causal inference and NLP.

Please give us a follow! @bleilab.bsky.social

1 3 20

Andrew Jesson @anndvision.bsky.social · Feb 17

youtu.be/X-Rgc7VzIp0?...

Andrew Jesson @anndvision.bsky.social · Feb 17

youtu.be/SJLe1UTqKvA?...

Nirvana - Lithium (Live at Reading 1992)

Read about Nirvana's 1992 performance at Reading here: https://www.udiscovermusic.com/stories/reading-nirvana/Listen to more from Nirvana: https://Nirvana.ln...

youtu.be

1