INDEX

Explanations

phrases that express strong emotion or emphasis

oai_token-act-pair · gpt-3.5-turbo

symbols or characters that denote a specific concept or idea

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 dock

-0.74

 oxide

-0.69

 bipolar

-0.63

 inspector

-0.62

 antip

-0.61

iciency

-0.61

 manned

-0.59

 sacked

-0.58

 reper

-0.58

 bonded

-0.58

POSITIVE LOGITS

wait

1.04

etc

1.01

WHERE

1.01

yet

0.99

there

0.99

well

0.98

they

0.95

literally

0.94

until

0.93

nothing

0.92

Activations Density 0.038%

phrases that express strong emotion or emphasis

symbols or characters that denote a specific concept or idea

No Comments

No Known Activations