INDEX

Explanations

words related to a collection of items or people, possibly in a list format

oai_token-act-pair · gpt-3.5-turbo

concepts related to boundaries or limits

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 giant

-0.62

 Toll

-0.61

 tension

-0.61

 weight

-0.60

 McGee

-0.57

 reuse

-0.55

erg

-0.55

 halftime

-0.54

 peanut

-0.54

 accent

-0.54

POSITIVE LOGITS

aries

4.97

ary

2.71

ARY

2.29

arily

2.11

naire

1.98

arial

1.94

arios

1.80

arians

1.76

aires

1.71

naires

1.59

Activations Density 0.010%

words related to a collection of items or people, possibly in a list format

concepts related to boundaries or limits

No Comments

No Known Activations