INDEX

Explanations

instructions or commands to follow

oai_token-act-pair · gpt-3.5-turbo

instances of the word "follow."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

pite

-0.82

inese

-0.78

Newsletter

-0.71

ãĥĨãĤ£

-0.71

cci

-0.69

risome

-0.68

 Scotia

-0.68

inished

-0.66

 ILCS

-0.66

urrection

-0.64

POSITIVE LOGITS

 directions

0.86

 closely

0.80

follow

0.75

 suit

0.75

ansen

0.74

itored

0.73

 blindly

0.71

 Follow

0.71

 behav

0.69

 obedient

0.68

Activations Density 0.036%

instructions or commands to follow

instances of the word "follow."

No Comments

No Known Activations