INDEX

Explanations

references to someone's actions or status

oai_token-act-pair · gpt-3.5-turbo

instances of the verb "was."

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

 entails

-0.75

inav

-0.74

 Previous

-0.67

 Which

-0.66

 Extend

-0.65

 Highlands

-0.65

 Highlights

-0.64

guy

-0.64

ð

-0.62

ESE

-0.62

POSITIVE LOGITS

 able

1.12

wolves

1.04

nt

1.01

wolf

0.97

 supposed

0.96

hes

0.93

 unable

0.92

 destined

0.92

 bitten

0.91

 afraid

0.91

Activations Density 0.336%

references to someone's actions or status

instances of the verb "was."

No Comments

No Known Activations