INDEX

Explanations

phrases indicating contrary or qualifying statements

oai_token-act-pair · gpt-3.5-turbo

phrases that convey a sense of caveats or qualifications

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

java

-0.71

ngth

-0.66

bid

-0.66

wine

-0.65

assic

-0.65

antha

-0.65

Discussion

-0.64

bis

-0.63

nesota

-0.63

utenberg

-0.62

POSITIVE LOGITS

 anymore

0.95

 anything

0.86

erest

0.83

 necessarily

0.81

 goodbye

0.74

ERSON

0.74

 anyone

0.73

 anybody

0.72

 exactly

0.72

terday

0.72

Activations Density 0.030%

phrases indicating contrary or qualifying statements

phrases that convey a sense of caveats or qualifications

No Comments

No Known Activations