INDEX

Explanations

phrases containing the word "interesting" and certain accompanying elements

oai_token-act-pair · gpt-3.5-turbo

expressions related to things deemed interesting

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

utz

-0.77

á

-0.76

oise

-0.74

redited

-0.72

©¶æ

-0.69

helle

-0.68

redit

-0.68

unker

-0.67

nery

-0.67

uts

-0.67

POSITIVE LOGITS

 interesting

1.06

 Interesting

0.89

 resil

0.88

 surprises

0.87

terday

0.83

 disse

0.82

 noteworthy

0.81

interesting

0.79

 juicy

0.79

tid

0.78

Activations Density 0.020%

phrases containing the word "interesting" and certain accompanying elements

expressions related to things deemed interesting

No Comments

No Known Activations