INDEX

Explanations

words related to directions or movement, specifically words such as "up" and "out."

oai_token-act-pair · gpt-3.5-turbo

directional words or phrases indicating movement or positions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

rouse

-0.67

 dstg

-0.65

 Turing

-0.63

TEXTURE

-0.63

 constitu

-0.61

hyde

-0.61

assic

-0.59

EStream

-0.58

tein

-0.58

 stim

-0.58

POSITIVE LOGITS

ward

1.02

stairs

0.99

coming

0.97

neath

0.90

ices

0.87

stream

0.85

numbered

0.85

look

0.84

raged

0.83

come

0.82

Activations Density 0.092%

words related to directions or movement, specifically words such as "up" and "out."

directional words or phrases indicating movement or positions

No Comments

No Known Activations