INDEX

Explanations

instances of the word "email", especially when used in relation to communication

oai_token-act-pair · gpt-3.5-turbo

mentions of emails and statements in communication

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

avorite

-0.85

estern

-0.77

rollers

-0.72

 idols

-0.72

tis

-0.71

utters

-0.68

rogens

-0.66

Ps

-0.66

rants

-0.66

 brakes

-0.66

POSITIVE LOGITS

 statement

0.79

statement

0.78

ority

0.72

 srfAttach

0.71

 standpoint

0.71

Statement

0.70

 thanking

0.70

STAT

0.68

ederation

0.67

 outlining

0.65

Activations Density 0.301%

instances of the word "email", especially when used in relation to communication

mentions of emails and statements in communication

No Comments

No Known Activations