INDEX

Explanations

short, interesting pieces of information or details

oai_token-act-pair · gpt-3.5-turbo

short, impactful statements or bits of information

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

issors

-0.76

 Voltage

-0.72

akespe

-0.69

Ú

-0.68

atchewan

-0.66

agne

-0.66

 Electric

-0.63

 warr

-0.63

Urs

-0.62

 Learning

-0.61

POSITIVE LOGITS

istically

0.94

geoning

0.90

istic

0.80

tid

0.79

eteenth

0.78

mouth

0.77

efully

0.75

ļéĨĴ

0.74

othe

0.72

ificent

0.68

Activations Density 0.058%

short, interesting pieces of information or details

short, impactful statements or bits of information

No Comments

No Known Activations