INDEX

Explanations

words related to physical attacks or aggressive behavior

oai_token-act-pair · gpt-3.5-turbo

references to various forms of assault

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

ãĤ©

-0.71

 snipp

-0.71

ocular

-0.68

FORE

-0.66

overed

-0.66

milo

-0.65

 Solitaire

-0.65

CFR

-0.65

ETA

-0.63

 darn

-0.63

POSITIVE LOGITS

 assault

0.84

uous

0.84

ments

0.82

ive

0.80

ively

0.79

quez

0.79

 perpetrated

0.77

iveness

0.77

entimes

0.75

 assaults

0.74

Activations Density 0.021%

words related to physical attacks or aggressive behavior

references to various forms of assault

No Comments

No Known Activations