INDEX

Explanations

countries and geographical locations, especially when mentioned in lists

oai_token-act-pair · gpt-3.5-turbo

the conjunction "and," indicating connections or lists of items

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

reads

-0.72

bo

-0.69

BIL

-0.69

Adapt

-0.68

Contents

-0.67

pter

-0.65

yout

-0.65

abuse

-0.65

matter

-0.65

inct

-0.63

POSITIVE LOGITS

 adjoining

0.88

 vice

0.84

 neighbouring

0.81

 neighboring

0.80

 vicinity

0.75

 onwards

0.74

 assorted

0.74

 other

0.73

 then

0.71

 elsewhere

0.70

Activations Density 0.333%

countries and geographical locations, especially when mentioned in lists

the conjunction "and," indicating connections or lists of items

No Comments

No Known Activations