INDEX

Explanations

words related to locations or geographical features

oai_token-act-pair · gpt-3.5-turbo

occurrences of the suffix 'ng' indicating ongoing action or a continuous state

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

è¦ļéĨĴ

-0.74

 absentee

-0.62

 consequ

-0.61

 Targ

-0.61

BIP

-0.60

balls

-0.60

footed

-0.59

toe

-0.59

 comet

-0.59

 Brach

-0.59

POSITIVE LOGITS

roups

1.21

roup

1.12

thening

1.07

allery

1.02

lasses

0.98

uay

0.97

raphics

0.96

oing

0.96

thens

0.96

lish

0.93

Activations Density 0.012%

words related to locations or geographical features

occurrences of the suffix 'ng' indicating ongoing action or a continuous state

No Comments

No Known Activations