INDEX

Explanations

dates written in a specific format

oai_token-act-pair · gpt-3.5-turbo

opening parentheses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.10.hook_resid_pre

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

 appropri

-0.73

 eaves

-0.72

 piping

-0.70

 paint

-0.70

 setting

-0.70

 registry

-0.69

 arbitrarily

-0.69

 discrim

-0.68

 haun

-0.68

 nuts

-0.68

POSITIVE LOGITS

Reuters

1.36

Photo

1.17

CNN

1.14

credit

1.11

Courtesy

1.09

AFP

1.08

UTERS

1.06

CBS

1.05

Filename

1.04

ANI

1.02

Activations Density 0.029%

dates written in a specific format

opening parentheses

No Comments

No Known Activations