INDEX

Explanations

words related to government surveillance, security, and communication interception

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

occurrences of the abbreviation "U.S." in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_post

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_post

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 challeng

-1.20

 enthusi

-1.11

 occas

-1.11

 nodd

-1.11

oun

-1.09

 mosqu

-1.08

 mathemat

-1.08

 exha

-1.07

 tiss

-1.07

 trave

-1.05

POSITIVE LOGITS

1.48

1.07

1.03

Va

0.99

$.

0.93

States

0.85

SN

0.82

0.81

Ns

0.80

Activations Density 0.029%

words related to government surveillance, security, and communication interception

occurrences of the abbreviation "U.S." in the text

No Comments

No Known Activations