INDEX

Explanations

phrases related to technical errors and instructions to try again

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "again"

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

ELD

-0.72

plates

-0.70

 Galile

-0.68

ties

-0.68

RIP

-0.68

ļé

-0.68

 Howe

-0.65

emies

-0.64

Ts

-0.62

Mot

-0.62

POSITIVE LOGITS

Cola

0.79

 recycle

0.76

 intrusive

0.73

 CTRL

0.66

 comma

0.65

 login

0.65

 Slug

0.64

 Scotia

0.64

 copy

0.64

ify

0.64

Activations Density 0.008%

phrases related to technical errors and instructions to try again

occurrences of the word "again"

No Comments

No Known Activations