INDEX

Explanations

the action of walking

oai_token-act-pair · gpt-3.5-turbo

instances of the word "walked" in various contexts

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

ional

-0.88

etric

-0.77

ency

-0.76

iled

-0.76

ãĥª

-0.75

usable

-0.74

relevant

-0.74

emin

-0.73

encies

-0.73

ccording

-0.71

POSITIVE LOGITS

 Walk

0.90

walk

0.88

 walks

0.84

 ashore

0.82

 stroll

0.81

 stride

0.81

 walked

0.79

 walk

0.78

Walk

0.77

escription

0.77

Activations Density 0.010%

the action of walking

instances of the word "walked" in various contexts

No Comments

No Known Activations