INDEX

Explanations

references to morning shows or programs

oai_token-act-pair · gpt-3.5-turbo

prominent television show titles and related phrases

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

wid

-0.64

illon

-0.60

USS

-0.59

hover

-0.57

 guiActiveUn

-0.57

eligible

-0.56

Ter

-0.56

igham

-0.56

quet

-0.55

icum

-0.55

POSITIVE LOGITS

 Practices

0.92

nered

0.88

 outweigh

0.84

ounters

0.82

 outwe

0.81

atoes

0.80

ornings

0.80

bye

0.70

anners

0.68

smanship

0.67

Activations Density 0.304%

references to morning shows or programs

prominent television show titles and related phrases

No Comments

No Known Activations