INDEX

Explanations

words indicating overt or overtaking actions

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "overt" and its derivatives, indicating situations or actions that are open and obvious

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

nesota

-0.92

UTION

-0.76

 Occupations

-0.65

SAR

-0.65

ĸļ

-0.65

haus

-0.65

©¶æ

-0.63

ERAL

-0.62

 Caps

-0.62

journal

-0.60

POSITIVE LOGITS

uned

0.99

aken

0.96

orial

0.95

orical

0.89

itled

0.88

imes

0.84

aking

0.82

urned

0.81

emin

0.80

isman

0.78

Activations Density 0.012%

words indicating overt or overtaking actions

occurrences of the word "overt" and its derivatives, indicating situations or actions that are open and obvious

No Comments

No Known Activations