INDEX

Explanations

names that contain "ha" with varying levels of activation, potentially indicating a preference for a specific name or concept

oai_token-act-pair · gpt-3.5-turbo

repeated occurrences of the substring "ha"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

papers

-0.87

atories

-0.76

rations

-0.72

ateur

-0.70

lace

-0.67

enhagen

-0.66

entric

-0.65

 largeDownload

-0.64

lines

-0.64

parts

-0.62

POSITIVE LOGITS

wn

1.10

iku

0.96

ha

0.94

pless

0.92

ichi

0.90

qua

0.90

jj

0.89

fter

0.85

pton

0.84

pp

0.84

Activations Density 0.010%

names that contain "ha" with varying levels of activation, potentially indicating a preference for a specific name or concept

repeated occurrences of the substring "ha"

No Comments

No Known Activations